Home

DISTRIBUTED DISPATCHING FOR EMBEDDED GENERATION

1. Last timestep s AC power All other currents Fig 7 4 Original inverter management set by user Selling orice Dispatched DC power Lookup REDMan table generator inverse Advertised AC power Dispatched AC power control Lookup table Desired DC power normal Fig 7 5 Inverter management with revised economic dispatching 113 power is limited by AC demand it will constantly buy more DC than the inverter is actually using The increment is small 5 watts in prototype but nevertheless it is an error Again a revised dispatching algorithm using incremental cost might well make this work around unnecessary Look up tables were also introduced in this revision The LabVIEW program is listed in Appendix H section H 2 3 7 3 Generator agent Unlike the previous efforts the agent for RE generation was a simple program The amount of power generated was determined by the RE source such as the amount of solar radiation or wind velocity The program only had to measure this power by reading the current and voltage and multiplying them together and then communicate it to the dispatcher using the standard communications routine In return it would receive the actual amount of power required Ideally it would operate a dump load to dispose of the difference between actual and required powers In practice it proved easier to arrange matters so that there was always guaranteed
2. PORTA W B 00001111 power F mode STATUS Z PORTA 4 finally saturate errc this is needed so the timer program can handle them nothrot movf bcf sublw btfsc goto movlw movwf errc W STATUS C TMAX STATUS C Skpsat TMAX errc don t execute this if we are tripped set multiplexing bit wait read 4 most significant bits into W mask put into power register Swap nibbles clear multiplexing output wait read 4 least significant bits into W mask OR with existing contents of power reg is throttle setting zero then cut out power stage can t be set to pump out interrupts faster than the is errc greater than TMAX if not skip the next bit if so let errc TMAX 201 reset the step index Skpsat clrf index toggle the interrupt edge trigger bit the next interrupt will come on the opposite edge bsf STATUS RPO our business is in bank one btfsc OPTION REG INTEDG if bit is set goto clear Clear it bsf OPTION INTEDG otherwise set it because it was clear goto tend and skip the next line clear bcf OPTION REG INTEDG which would just clear it again tend bcf STATUS RPO back to bank zero wreck any interrupts that might have happened meantime bc INTCON INTF Clear zero crossing interrupt bc INTCON TOIF Clear timer interrupt send zero to PORTB since this code runs instead of wave step handler cl
3. bcf STATUS RPO bank 0 clear output latches bsf STATUS RPO bank 1 movlw B 00000001 all outputs TRISB 0 except RBO movwf TRISB bsf OPTION REG NOT RBPU turn off internal pullups bsf OPTION REG INTEDG interrupt on rising edge bsf INTCON INTE enable as interrupt source Start interrupts and we got ourselves a convoy bcf STATUS RPO bank 0 bsf INTCON GIE global interrupt enable goto main e ck ck ck ck ck ck ok kk kk kk ck kk kx kx KKK MAIN CODE e ck ck ck ck ok ck kk kk kk ckockckockokockockockck this doesn t need to do anything all the work is done by interrupts Having the processor in an endless one instruction loop might increase interrupt latency because goto takes 2 cycles so I give it several hundred pointless instructions main nop no operation org LUTBASE 1 then execute a stack of empty memory goto main loop forever phoo ee END A 19 References Personal communications with Trace Engineering and Statpower 2000 2 P929 Recommended Practice for Utility Interface of Photovoltaic PV Systems IEEE Sep 1998 3 Power MOSFET Design Taylor B E Wiley 1993 various 4 Power Electronics Lander C W McGraw Hill 1993 pp 198 216 5 Alternative Transients Program website Michigan Technical University http www ee mtu edu atp 2001 6 Development and Testing of an Approach to AntrIslanding
4. eee entree ed re erre EYE EC EYE TE Ree E eee Reed 220 D 5 CONCLUSIONS x eae ere ODE EE P ER Fee aa e E ERE dede e UR ER 221 APPENDIX E MODELLING INVERTERS IN ESP R eere en eneses sess een snenenss se senenesss 225 E 1 ABOUT STRUCTURE OR ESP R 4 oet ee URBE e RE 225 2 THE POWER FLOW SIMULATOR PES ens 226 3 ACTIVE CONNEGTIONS teet e e UB RR a ee ACO RSE 226 E A ADDING NEW POWER 228 E 5 RESULTS ets ut FARO HEREIN E 6 CONGEUSIONS iino DETUR er DIU en ERROR ERA ACKNOWLEDGEMENTS E 8 REFERENCES eer ORUM pO UN deu nd RDUM APPENDIX HYBRID PV SYSTEM CONTROLLER eee tn tanta ta tanta tota a tato ta natn 231 E ABOUICTHESYSTEM iR REA ERA ES T RE RE NE EE 231 F 2 CONTROLLER REQUIREMENT Sveiser riesi iernii tri Se Ea ES E AET E OaE E 231 F 3 DDESIGNIGONSIDER ATIONS 231 F4 THEORY OE OPERATION 5 oet ete d ce EROR GER ERO Ote E TENES 232 E CIRCUIT DESCRIPTIONS a REVO GE UR GEO ER ERU DEREN 232 E 5 I Gontrollernodule 232 F 5 2 Power supply metering module esee tette tette tenente eene 233 E 5 3 y stem WIN aco EUR HER e RUN Ci 233 F 6 MODIFICATIONS 5 ER WR RU Ee e C etat ete etn 234 APPENDIX HOW LABVIEW WORKS ccsssssssssscsssscssssssssssssssses
5. resin esiste etatem enean tato tenen ta 18 3 1 HOW DOES CONTROL APPLY TO POWER SYSTEMS ennt inneren 18 3 2 LOAD FREQUENCY CONTROL OR RARE 19 Simple gOVerfiOraxc cose Se eR b e 19 3 2 2 Implementation in large networks eese eterne nente nenenene 20 3 3 AUTOMATIC VOLTAGE REGULATION easar e a e tes tetas 20 Sided Fundamenials sc o o NINE DERE OR RR 21 33 2 HOW AVR AS Set Up cR ERR EN A tds 22 3 3 3 AVR considerations i E 22 34 PROTECTION noto eO OU HR RR a IO 23 3 4 1 Types of protection eee 23 DAZ Pr tection JOE G 24 3 4 3 Anti islanding protection eese entente tenente tenete nnne 24 3 5 IMPACT OF EG ON EXISTING CONTROL ener tentent tenens en 26 3 6 ECONOMIC DISPATCHING c ccccscssescssessesessescssessscssescsscsssecsessessscaesecnesassecsessesscaesssaescaesesaceesesseeesaeseesees 26 NONI G38 ve 26 9 6 2 Application to EG ie ert istae ts 28 AE niri aca WE ca ea E REEE 28 S LAL Whatexactlyis stability2 et ER RU E ee te 28 3 22 Classical stability studies e ed eee epe e 29 3 73 Application t
6. G 7 Multitasking LabVIEW is a multitasking system First of all t is possible for several parts of the same VI to be running simultaneously Indeed if one part of the block diagram is disconnected from the rest and so is not told when to execute by dataflow it will execute in parallel with the rest of the diagram This can be useful as an easy way of writing multithreaded programs However it can also be a source of confusion The dataflow model is useless when communicating between multiple threads To work around this LabVIEW provides some functionality similar to traditional such as variables of global local scope semaphores etc Unfortunately it is easy to get into a tangle when using these alongside dataflow programming whose order of execution is not always obvious It is also possible for several VI s to run simultaneously on the same computer They may be run independently with each preserving its own environment or they may share data through global variables G 8 Modularity Alternatively one VI may be placed inside another analogous to a subroutine in a conventional program As might be expected it is termed a subVI It is not immediately obvious how a program that is operated by clicking on buttons on its front panel can be turned into a subroutine In practice this is handled by giving every VI a second shrunken version of its front panel termed the icon The icon sports terminal
7. could have been caused by actual changes in line voltage to an extent More annoying was a slight zero offset 3 6 watts that could not be eradicated It was eventually tracked down to crosstalk the signal on the voltage channel was bleeding into the current channel This was happening inside the soundcard circuitry itself and not very much could be done about it short of upgrading to a better soundcard In any case this was an error of under 196 of f s which was actually better than some commercial power meters Figs 13 5 and 13 6 are sample screenshots from the software 13 5 shows the power analysis of a 60 watt light bulb The power is correct to within a few percent and the error could be due to the bulb rather than the measurement system The only problem with this measurement is that it shows a small amount of reactive power whereas a lamp is obviously a pure resistance This would tend to suggest a slight phase error between channels or it could be a result of capacitive inductive crosstalk between channels 13 6 shows a highly reactive and load actually the AC power unit of laptop computer This example illustrates the need for such complex means of measuring power By using a voltmeter and ammeter here and multiplying the voltage and current readings the result would be similar to the apparent power VA rating which in this case is 47 However the real power measurement shows that the unit is onl
8. different markets the mathematics for economic dispatch could become daunting In renewable generation the short run marginal cost is zero since the fuel is free The long run marginal cost may be assumed constant which is not a problem from the ED viewpoint rather it simplifies matters greatly 3 7 Stability 3 7 1 What exactly is stability Stability is rather loosely defined in power systems work Basically when all of the control systems discussed above are set up properly and working in harmony stability is the result If they are wrongly configured or interact in unforeseen ways then the grid may become unstable For example if load following control is not working properly there may not actually be enough reserve to meet a sudden increase in demand The instability that results is known as frequency collapse a nightmare scenario involving the whole national grid The remedy for frequency collapse is drastic demand has to be shed until the frequency starts to recover To take another example if the protection system is not properly configured then it is possible for a fault to trigger a chain reaction which trips out far more equipment 28 than was necessary to clear isolate the original fault This may also be thought of a form of instability In a similar way a fault in the transmission network can suddenly upset the flow of real power The resulting transient can cause synchronous generators to pull o
9. e Tine ien qn eii eie cie pe ie a ebat is 115 T 5 DEMAND AGENT RR RE 117 ZO Theoriginial a bia qi eee o ta b etas 117 118 7 5 3 Prototype power measuring 118 PO SUMMARY n EA NIRE 122 Ti REFERENCE S OR EN 122 CHAPTER 8 EXPERIMENTS eere tenen nente tentata senes tonta 125 8 1 SUMMARY OF DEVELOPMENT WORK SO FAR eese 125 OUI T SOftWake s eet eet ederent dere Reed 125 0 4 2 First hardWafess i e s e ER ER t 126 0 13 Mark Two hardWake e eee dO Rte ER 126 SLAC Mark TWO SO IWAGEG 4e t aaa 126 8 2 EXPERIMENT DESIGN eR TRE ERR TREINTA 127 8 3 7 TESTING DISPATCHER 128 8 4 CHECK AGENT PROGRAMS ASSUMPTIONS ccccscssesssscssescsecsesscaesecsesceesecsessesescaesssaeeesesaeceeseseeseeacaeens 129 8 5 TESTING COST MODEL ives RE ten Heo E IO EIE E EE ens 129 SD TL Inyerier cost models stein ie ERAT sin savin abe o aot 130 6 5 2 Battery cost model sss cada ten 130 6 53 130 8 6 EXAMINING MULTIPLE DOMAINS eese eret 130 8 7 OBJECTIVES OF EXPERIMENTS tnr ei ne ree rete ee es 131 8 87 SUMMAR Yo ette im Re eee Ne
10. Eco restructuring Implications for sustainable development ed Ayres R U United Nations University Press 1998 Online at http www unu edu 2003 Wind on the System Grid integration of wind power Renewable Energy World Vol 3 No 2 March April 2000 pp 61 71 Embedded Generation Jenkins N et al London 2000 259 17 Chapter 3 Power systems control The need has been identified for a dispatching system that can control large numbers of electrical generators without being excessively complicated In order to have a better idea of the requirements inherent in such a system it will be wise to review the control engineering aspects of power systems to get a good impression of the issues involved This section is a brief review of power systems control the techniques used the issues involved and the technologies employed with a particular bias towards embedded generation The reader is assumed to be familiar with some fundamental concepts of control engineering itself such as feedback and stability See e g 1 3 1 How does control apply to power systems Control engineering is fundamentally concerned with applying a self correcting or self regulating tendency to systems that do not naturally have such a tendency This concept is applied in many ways throughout the whole electrical power field However in the sense of interest at present power systems control essentially means keeping
11. HEA LON PEAS ZL 960 onsejd doyoedes Ajajes x sse ayeb pueu JojeJeduio2 penb paads BEEWT en duie do jenp Ajddns ajbuis ssc Wi juauiu yoy 00 20 4 87 59 001 png 1 L r3 5 to D AS 21 A ts gic circui lo ing PWM generator and switch Fig A 9 169 Output 5 S5 S6 54 10 time ms Fig A 10 Switching circuit and simulated waveforms Carrier frequency reduced for clarity 170 These rules be turned into simple collection of digital logic gates by using well known techniques Firstly the inputs to the circuit must be defined they are P the PWM pulsetrain and G the polarity of the grid voltage G 0 when grid is negative and 1 when it is positive Let the outputs be S1 S2 S3 S4 S5 S6 and let a 1 indicate that the corresponding switch is closed A truth table can then be drawn In the interests of clarity this will not be done in the conventional manner instead of the usual 0 or 1 the outputs can also be P or P Table A 1 From this it is easy to write Boolean expressions for each output S1 PG Eq A 1 S2 P G Eq A 2 S3 PG Eq A 3 S4 P G Eq A 4 S5 G Eq A 5 S6 G Eq A 6 Before proceeding from here to an actual logic circuit there is one detail that must be taken care of S1 S2 are in se
12. With these caveats in mind the rules of the game can be formalised not to mention proposing an acronym for it Rea time Embedded Dispatch Manager or REDMan 4 4 REDMan proposed The REDMan system is thought of as consisting of sources of power demands of power a central dispatcher and a communications system which allows messages to be passed between them Physically it will probably consist of several computer programs agents for sources and demands plus a dispatching program running on one or more computers or embedded controllers interconnected by a network of some kind Within this framework it may be envisaged working as follows 4 4 1 Algorithm and communications 1 A source must advertise to the dispatcher the maximum amount of power it can supply and its per unit price The price is allowed to be zero or even negative The source will then be informed of the amount of power required of it and must then supply that amount The amount of power required may change at any time and may be any amount between zero and the advertised maximum If the source 43 requires to change the maximum available power or price for any reason the advertisement should be repeated In advance of switching on a demand must state the amount of power required to fulfil it and the maximum per unit price which it is prepared to pay for the fulfilment It will then be informed as to whether or not it can be fulfilled at that price If
13. fine movf tmrtmp W bcf STATUS C sublw TREF btfss STATUS C test TREF TMRO goto fneg if positive or zero incf errf F increment it btfss errf 7 if error is now greater than 7F goto fprot skip rest of routine if it s not clrf errf reset it to zero incf errc F and increment coarse error goto fprot end of if zero or positive code else if negative fneg moviw 1 bef STATUS C subwf errf F take 1 away from errf btfsc STATUS C has it rolled over from 00 to FF goto fprot if not skip next instruction decf errc F decrease coarse error now timer calculations are finished test if frequency in tolerance This is done by checking the timer preload errc which of course depends on the frequency Now there are 128 steps in each cycle 100 cycles per second and the timer runs at 2 457 600 ticks per second So we d expect there to be 192 ticks in each cycle But the timer interrupt has a latency of 14 ticks etc Fear not the value which the timer should take is calculated in the defines it s TREF This routine is vaguely based on Sandia Labs SFS protection system increase frequency violation counter every time errc strays outside tolerance rot movf errc W bet STATUS C sublw TLO btfsc STATUS C is errc less than TLO goto finc movf errc W bcf STATUS C sublw or is it more than THI 200 btfss goto goto incfsz incf goto finc else fok movf btfsB decf
14. if the counter hits movf bcf sublw btfsc goto ftest bsf movf occurred sublw btfsc goto movlw movwf movlw movwf STATUS C finc fok fvi W fvi F ftest fvi F STATUS Z fvi F fvi W STATUS C CTS STATUS C fnotrip PORTA 4 pstat W TRIPPED STATUS Z fnotrip TRIPPED pstat CTR fvi skip if it is not test if this will roll over if it won t then increment Skip decrement if frequency was OK decrement fvi Unless it s zero move fvi to itself in order to test for zero CTS Bring the show down Skip if fvi greater than CTS branch to no trip if fvi within limit otherwise immediately cut out power stage test if status tripped ie trip has just because we only want to reset fvi once and not every time we do this test which would lock us up for good Status tripped reset fvi if the counter hits 0 Start up again fnotripmovf btfss goto movlw movwf fvi W STATUS Z ttest RUNNING pstat next instruction skipped if fvi 0 set status to running now get the new throttle setting for this cycle 8 bit setting multiplexed in as two lots of 4 bits due to shortage of IO pins ttest movf sublw btfsc goto bsf nop nop movf andlw movwf swapf ber nop nop movf andlw iorwf throt simple sleep btfsc bsf pstat W TRIPPED STATUS Z nothrot PORTA 4 PORTA W B 00001111 power power F PORTA 4
15. lines Ducted wind turbine 100W nominal University of Strathclyde experimental 6 4 2 Structure The system was configured according to Fig 6 1 The PV panels are connected as two 24 volt arrays and brought to the 24 volt battery bank composed of two 12 volt batteries in series The two ducted wind turbine DWT modules are also connected to the battery bank Since they generate AC a rectifier is used with each The inverter takes its input from the battery bank and feeds its output into the electrical grid It is plugged into a wall socket like any other electrical appliance AII voltages and currents of the DC system and also the inverter status are logged controlled by a computer with DAQ card Via Ethernet this computer connects to another computer which is physically remote This second computer monitors controls the status of the loads consumers of AC power 82 n 2 ducted wind turbines 12V 2x BP Saturn PV modules Rectifiers 2x Solarex MSX 50 PV Modules Roof 2 x 12V lead acid batteries Ethernet 6th floor Office area gt Demand sensors Load management computer Assorted loads Fig 6 1 Layout of the test bed 6 4 3 Layout The proper siting of the various parts of the system required some thought Firstly and most obviously the PV modules would have to be sited outside and high up where they would not be shaded by other structures A s
16. such as these voltage measurement is almost useless The final and most sophisticated method is the charge meter The heart of this system is an electronic coulomb meter connected in series with the battery which measures the actual quantity of charge entering and leaving The coulomb meter works by measuring current and taking the time integral Of course where integration is involved there is always the possibility of errors building up to an unlimited extent and so this method relies on a secondary means of resetting the coulomb meter to a known state Commercially available systems use a voltage measurement to do this when the terminal voltage exceeds a certain amount the battery is considered to be full and the meter is reset 5 The charge meter method is applicable to all known battery chemistries since every chemistry shows a rise in voltage during charging once the battery is full In its basic form it takes no account of the capacity s dependence on discharge rate and battery ageing and is usually calibrated in amp hours out the number of amp hours extracted from the battery with reference to its fully charged state The user is left to estimate what fraction this represents of the available charge Some commercial charge meter products 6 provide a readout in percent remaining or miles of range for electric vehicle applications and it might be inferred from this that they have proprietary ways of compensating for ca
17. 7 1 4 Factors affecting battery voltage Unfortunately just as the capacity varies from its nominal value so does the battery voltage A single lead acid cell nominally gives 2 volts But when the cell is almost discharged the voltage can be only 1 6V and when it is reaching the end of a charging cycle it can rise as high as 2 4V Thus a 12 volt lead acid battery can give anything between 9 6 and 14 4 volts This change in voltage is brought about by three mechanisms First is the polarising voltage The chemical reactions in the battery being irreversible require energy to drive them They draw this energy from a small voltage drop associated with the reaction The size of this drop depends on the current drawn the state of charge and the temperature The polarising voltage is a log function of the current and is zero when the current is zero Second even in the absence of any current the open circuit voltage depends on the state of charge and the temperature The dependency is approximately linear in both voltage and temperature Thirdly the electrical resistance of the battery causes a voltage drop proportional to the current The internal resistance is usually very small and so this component can often be ignored The voltage measured at the battery s terminals is the sum of open circuit voltage polarising voltage and voltage dropped across the internal resistance and so varies with state of charge current and te
18. How lead acid batteries work aiia 93 7 1 3 Variation of capacity with discharge 93 7 1 4 Factors affecting battery 95 Fil De Ageno andservicelife i ee eee 95 7 1 6 Calculating energy content not 5 96 7 1 7 simpler battery management model eese 96 1 7 9 Taking careof batteries ue tene 98 PAO USING ihese deas te ee eee eee 99 7 1 11 A model for battery wear and 100 7 1 12 Practical implementations sine e eese tette tenete tnter nennt tenentes 102 7 1 13 Practical 104 A 1Jd4 REDMan connection soe et d dt de e etsi 108 V5 Lesung BatMan uid eo ee enia tte eR 108 1 2 INVERTER AGENT ert 110 INT DAI EE 110 152 2 HER RUE ERR BU NRI He REN EM eS 110 7 2 3 Calculating inverter efficiency eese esiste teinte tette tette tete te tete tenete tenente 111 7 2 4 Simple fixed price algorithms eese estere tette 111 7 2 5 Economic dispatch Mark One 112 7 2 6 Economic dispatch Mark 112 3 3 114 7457 DUMP LOAD AGENT idet ERR ERE ERREUR 114 7 4 I Implementation use aee tester deerat ede 115 7 4 2 Performance 5
19. RF oscillations a common problem in MOSFET circuits D1 senses the DC bus voltage via divider network R9 R10 R11 R17 When the voltage at D1 input exceeds the internal 2 45V reference D1 sinks current turning Q1 on which sources extra current into the MOSFET gates increasing the gate voltage and turning the devices on Circuit values are chosen so that the MOSFETS can be turned fully on by the regulator even when the control voltage input is zero Potentiometer R9 allows the regulating voltage to be set between 24 and 30V Cooling fans not shown in circuit are connected in series for 24 volt operation and connected across R1 R2 so that fan speed depends on current draw 207 B 3 Computer interface The control voltage was derived from 8 digital output lines on the computer using an R 2R ladder type DAC borrowed from a prototype inverter This has rather a high output impedance and the dump load control input will load it down somewhat Fortunately the input turned out to be sensitive enough that this was not a problem B 4 Construction The apparatus was assembled on a large heatsink 200x250x50mm with a bracket to hold the cooling fans in place An aluminium casing was made to protect the wiring and terminals from short circuits and to allow the assembly to stand upright for optimum convection cooling This was an issue because the fans did not actually start spinning until about 80 90W dissipation was reached 208
20. This was done because it was known that the RE generation only 30 watts on average would not be sufficient to cover the load and also to test economic dispatching and give the inverter management more changes in power to deal with The system ran satisfactorily for four days with no obvious problems At the end of this period the data files were collected for analysis Fig 9 1 shows the dispatched powers in the AC domain over the entire test run period This gives an overview of the system performance 133 50 40 N 10 06 80 75 09 70 95 09 00 85 09 06 17 09 91 57 09 51 57 09 80 7 09 70 67 09 00 19 09 06 79 09 91 99 09 21 89 09 80 00 15 70 20 15 00 70 19 06 20 19 91 60 19 21 16 80 51 19 70 91 19 00 81 15 06 mm ss Time hh rformance of system ing pe tch ispa D 1 ig 9 F 134 9 2 Energy balances The first task was to check that all the powers in the dispatching engine added up This was done by summing all the supply powers from the test run log file summing all the demand powers and taking the difference Results are summarised below for each domain Energy demanded Wh Energy supplied Wh Difference Wh 1513 205 1513 194 0 011 AC 4669 583 4669 583 0 000 The AC numbers are exact however the DC numbers are very slightly out Whatever Table 9 1 the reaso
21. dumpload etc are performing Fig 9 3 shows this and it 1s rather a different story to the dispatcher s point of view The PV power is the same but the dump load is not living up to the performance seen by the dispatcher It kicks in late and in a half hearted manner dissipating only 5 watts instead of the 15 it is commanded to The remainder goes into the battery which swallows it without any appreciable effect on the voltage This suggests that it is not full otherwise the voltage would rise The S O C at this point is in fact only 97 5 which is unusual considering that the batteries started at 100 and that they were not supposed to be delivering any energy at all The slow operation of the dump load can be blamed on the dump load controller program itself As discussed earlier this uses a simple integral controller to adjust the current passed until it is equal to the desired value Unfortunately because of the square law characteristic of the transistors in the dump load the integral gain must be chosen to give stability at the highest currents where the square law gain is highest This makes for a very sluggish response at low currents However the unexpected behaviour of the battery is puzzling and requires a detailed investigation 9 5 3 In search of missing energy Fig 9 4 shows the battery parameters over the duration of the experiment The dispatched battery power is as expected mostly zero except for a few chargin
22. frequency common mode signal This seemed to be a major problem Solutions like opto isolation were investigated and ruled out on grounds of complexity or inaccuracy It was finally solved by a modification to the switching circuit See Fig A 8 It works because the full bridge as it is usually set up has only one side driven with HF at a time The sides change roles with every half cycle The inactive side has the lower switch turned on continuously in effect earthing the un driven end of the load So two extra transistors were added solely for this function Instead of taking the load current directly to earth they were arranged to divert it through a current sensing resistor Crucially one end of this resistor is now earthed solving the common mode interference problem A simple op amp differential amplifier finishes off any remaining common mode due to voltage drops in the ground paths 166 Fig A 8 Modified full bridge inverter for easier current sensing 167 A 8 Control circuits A 8 1 PWM generator This is perhaps a misleading title because the bang bang method of control causes an oscillation involving the entire circuit So it is hard to say just which part is actually the PWM generator For the purposes of this discussion though it will be defined as the part from which the PWM waveform first emerges which means the comparator circuit which compares the modulating wave to the measured current The
23. however and computing power is increasing 30 rapidly all the time Great steps have also been made towards a theoretical analytical understanding of the issues described above See 17 3 7 4 How will EG affect stability Due to the difficulties in modelling and intuitive understanding discussed above it is unfortunately quite hard to say At one extreme if it were controlled in a totally optimal manner it might be as good as or better than centralised power systems of today At the other extreme through blind projection of the technologies and policies currently applied to EG it could wreck the network completely No one really seems to know how large numbers of EG units each incorporating its own LFC and or AVR will behave when they are connected together and to a network that also contains other control algorithms reactances and rotating machines Because the potential risks are so high it is understandable for power systems engineers to take a conservative point of view They cannot prove that EG will integrate satisfactorily and so cite Murphy s Law 18 to prove that it will turn out to be disastrous However this viewpoint must be weighed against considerations of redundancy which suggest the opposite A detailed review of reliability theory is outwith the scope of this work but broadly speaking small generators are simpler than large ones and so may possibly break down less often Therefore considering a singl
24. might not be economically desirable An alternative line of attack might be to ensure that islands never drift out of synchronism at all very accurate timebase to within a few nanoseconds can be derived from Global Positioning System GPS receivers If enough generators in an island were locked to the GPS timebase and the generation in the rest of the network was also locked then there would be no problem The price of GPS receivers 1s currently around 150 and so this standard is accessible even to relatively small EG units 25 The final objection poor power quality is probably the most complex and demanding of all The network may island in a vast number of ways in theory it can split at any point where a fuse or circuit breaker is present There is no way of knowing whether generation will match demand within whatever islands might form The best approach here is probably to maintain tight voltage and frequency trips on all embedded generators In this way customers should either receive power of acceptable quality or none at all So it can be argued that there is no fundamental objection to islanding It is simply a matter of difficulty and expense to manage it properly would mean extending the advanced protection and switching techniques used in HV transmission into any part of the network that might island which would entail fairly radical changes to the distribution equipment In the short term of course it may prove
25. not yet a firm idea of how long this will be or whether it will be synchronous between different domains Perhaps there is no reason why it should be and indeed it may well be more robust and easier to build if it was not When considering the required length of timestep it is also important to remember that all contracts to buy sell power in the system implicitly last for one timestep Therefore if a source agrees to sell power it must commit to maintaining the required power output until the end of the current timestep Obviously if a renewable generator is to do this it must store some energy since it cannot guarantee that the renewable energy flux will not fall during the commitment period Since renewable energy systems require storage anyway having storage distributed along with RE generators might be desirable However the longer the timestep the larger the storage required As a first approximation the timestep should be short enough so that the maximum expected shortfall or excess of RE in the course of one timestep is small compared to the storage available 4 4 3 Timestepping and commitment The above analysis assumes that each commitment is perfectly strict and immutable In practice it may be better to build in some freedom for manoeuvre It may also be undesirable to have demands constrained to units of one timestep unless the timesteps were made very short Perhaps the average power delivered over a certain number of t
26. that the valve in this scenario is virtual and so has no power losses Of course it is unlikely that generation could be made to track the changes in demand instantly and so there might well be transient feed back of power However it is hard to object to this on legal grounds because there are many loads currently in use which do actually feed power back into the supply from time to time It is standard practice in 74 industrial motor drives where it is known as regenerative braking As the name suggests when the drive is to be braked the motor is reconfigured to act as a generator The best known application of this principle is in lifts elevators where it allows the potential energy of the cage and counterweight to be recovered instead of wasted It also finds application in drives for rotating machinery with very high inertia that needs to be stopped or reversed frequently for example in rolling mills In any case condition 2 specifies only that no energy should be fed back into the point of interconnection In the present case this would be the connection between the University building and the power company s substation It is hard to imagine that a small scale experiment could ever feed in enough power to overcome the demand of the whole building even in the middle of the night 6 2 Generating plant The choice of generators is mostly dependent on what equipment is available It so happened that there were a numb
27. the like However by careful design it should still be possible to extract useful results 4 7 Conclusions A proposal has been advanced for a new means of power systems control called embedded dispatching which may be suitable for large numbers of embedded generators A possible protocol for transferring the minimum of information necessary for distributed dispatching and a possible algorithm for performing the dispatching were described These systems have been kept simple by making assumptions about the behaviour of the network and requiring control systems to be incorporated in generators and loads such that those assumptions hold true Arguments in favour of these assumptions were presented but it will be vital to back them up with evidence from experiments or simulation The next chapters set out to do just that starting with an implementation of the REDMan dispatcher in computer software 51 Chapter 5 Developing the REDMan software So far a new system for embedded dispatching has been proposed and arguments made in favour of a real experimental test of it Now it is time to make a first attempt towards constructing that experiment A reasonable place to start will be with the combination of the dispatcher and the communications network that connects it to the sources and demands it manages The goal here will be to implement these as computer programs with the capability of interfacing to real world sources and deman
28. unless you satisfy either of these two conditions 1 Your installation is completely disconnected from the power company supply before it is connected to the alternative supply and stays disconnected In other words an either or arrangement 2 You obtain the written agreement of the power company before connecting any power source of your own in parallel with their supply Condition 1 could be satisfied by means of a transfer switch on the premises where embedded generators are used so that loads may be powered either by the EG or the grid whichever is available or most economical The switch is mechanically interlocked so that the EG can never be connected to the grid While this system provides the highest degree of protection against islanding it is very restrictive from the point of view of economic dispatching because the EG must be sized to meet the maximum expected demand There are also problems with operating the transfer switch while loads are drawing power since this is liable to cause spikes and or brief interruptions of the power supply Really this mode of operation is best reserved for backup power applications where the transfer is only made in case of a power failure Another possibility would be to actually connect the EG equipment to the grid but controlling it in such a way that supply never exceeds onsite demand and hence no export to the grid is possible This is rather like the one way valve concept except
29. 5 So T di 9 dt 4352 T dt Eq A 8 Substituting 12 L for di dt Eq A 7 and performing the integration 9 3 5 9 thus t 0 29L and f 1 2t 1 72 L so to have 30kHz L 1 72 30000 57 So the requirement is for two 30 H inductors to handle 71A without saturating Also their combined DC resistance must not exceed 0 006 The options were somewhat limited it was necessary to use off the shelf cores because having custom magnetic assemblies made would be too expensive and time consuming The largest ferrite available was Ferroxcube s ETD39 and for each inductor two of these core assemblies were stacked to double the core area These were wound with a 5 turn coil made of 5mm dia copper pipe with 0 7mm wall thickness and assembled with an 0 8mm approx airgap between core halves 184 With these coils the switching was somewhat faster than ideal 70 100kHz This suggested that the inductance was too small However since the inverter did not seem to be suffering from excessive switching losses and larger inductors would have been a major problem requiring custom made ferrite assemblies it was decided to use them anyway A 12 Assembly and snagging Most of the circuitry was put on a PCB which was a modified version of the board used in the Mark One There were three main revisions First was an extra negative supply generated by a small DC DC converter for the current sense amplifier
30. C MMy Documents labview backup MlabviewNlabviewNphdNrealthingNrm load close vi invcon vi Documents labview backup labview lLabview phd demo invcon vi sysglobal vi Documents labview backup labview labview phd realthing sysglobal vi investimate2 vi Documents labview backup labview labview phd realthing investimate2 vi investimaterev vi Documents labview backup labview labview phd realthing investimaterev vi 259 Be Go Gel A S a 260 H 2 4 REDMan dispatching server server3 vi server3 vi Connector Pane Front Panel 261 Number of generators I 032 Datasocket URLI PATH abe GEN gt Generator lopen a pair of datasockets Number of loads each generator and load lone going each way is a local variable yeeuk close all the sockets when wel lare done lopen and initialise the control port aa Block Diagram 262 Hidden parts of block diagram 1590505000009 00 RES Tr Doo 1 0 4 He 0999595 i 09599595 ZEE 2 0 4 FEE 5995 5519597575750 Ap Write out the solution from previous timestep f Use control port to tell clients they can go now clients your time is up Status writing out solution H ina ch H x
31. Control Error ACE A controller algorithm is used to adjust the output of the generator s in that area in order to regulate the ACE to zero Many different ways of deriving the ACE and many different control algorithms such as ordinary P I D fuzzy logic etc are described in the literature e g 5 6 3 3 Automatic voltage regulation The frequency of alternating current is the same everywhere in the grid but not so voltage which varies from place to place Transformers step it up and down and the resistances and reactances of cables and lines cause the voltage to drop or sometimes even to rise depending on the current passing The purpose of AVR is to make sure that every consumer gets a voltage within the statutory limits at all times 20 3 3 1 Fundamentals AVR systems are fundamentally the same as any other control system The voltage at a point in the network is measured compared to a desired value and the error signal actuates some kind of voltage adjusting device in such a way as to extinguish the error There are a few different types of voltage adjusting device One very widely used type is the tap changing transformer This is like an ordinary distribution transformer but has a high speed switch mechanism that changes the tapping point on one or other of the windings according to the instructions of the AVR system 7 8 Another way of adjusting the voltage is by reactive power compensation 9 This works
32. H Polling clients wait for clients to gc WHeHsHisHsHeHeHsHsHsHeHeHaHsHeHeHeHsHsHeHeHsHuHeHeHsHsHeHuNsHsHsHeHsHsHeHeHeHsHsHeHeHsHeHeHeHsHsHeHeHeHsHeHeHeHsHeH List of SubVIs DataSocket Open Connection vi E C Program Files National InstrumentsNLabVIEW 6 vi lib platform dataskt 1lb DataSocket Open Connection vi pen 2 DataSocket Write Double vi nen C Program Files National Instruments LabVIEW 6 vi lib platform dataskt 1lb DataSocket Write Double vi 24 DataSocket Write Boolean vi C Program Files National Instruments LabVIEW 6 vi lib platform dataskt llb DataSocket Write Boolean vi C Program Files National Instruments LabVIEW 6 vi lib platform dataskt llb DataSocket Close Connection vi os DataSocket Close Connection vi 263 iy Ny 5 a a m N m or TU aj 7 el DataSocket Read Double vi C Program Files National Instruments NLabVIEW 6 vi lib platform dataskt 1lb DataSocket Read Double vi DataSocket Update Data vi C Program Files National Instruments NLabVIEW 6 vi lib platform dataskt 1lb DataSocket Update Data vi DataSocket Write Integer vi C Program Files National Instruments NLabVIEW 6 vi lib platform dataskt 1lb DataSocket Write Integer vi pathgen vi Documents labview backup labview labview phd demo pathgen v
33. In Japan Vol 136 No 3 2001 Beyond POTS Electricity generatio n selling control Johnson C IEEE Power Engineering Society Summer Meeting 2000 Vol 1 pp 585 588 72 Chapter 6 Building a testbed It is now time to take the step of hooking the dispatching server up to a collection of actual electrical generators loads and storage systems This could be quite a difficult process and there are a few main problems which can be foreseen in the light of previous discussions The first problem will be finding suitable generating equipment given the constraints imposed by the scale of the project This could well rule out heavy machinery like fuel cells and IC engine powered generators which tend to be costly spread noxious chemicals and fumes around and require large amounts of care and attention The second probkm will be setting up an interface between the generating plant or indeed the electrical loads and the computer based dispatcher The generating plant must be able to accept commands such as change output power to 300 watts Similarly the loads should be capable of being switched off or dumped by the dispatcher A generator or load which fulfils these requirements is sometimes called dispatchable At the review stage it was found that dispatchable small generators and dispatchable loads are quite rare The third problem is to manage grid co operation It is a very attractive proposition to use the exis
34. It was immediately obvious that the inverter could not meet its rated power spec although it had been designed for 600W it proved impossible to get any more than 570W of it and this accompanied by some very sinister crackling noises coming from the filter chokes A 550W limit was set on subsequent tests to reduce the possibility of any damage until the problem could be found As tests continued it became obvious that it would not quite meet the efficiency spec either the efficiency at 550W was only 86 The peak efficiency was 90 at 260W output Harmonic distortion of the AC line current was somewhat higher than spec too with a measured 5 8 at 550W There was also a non linearity of power with power control The power began to compress at higher levels and it seemed likely that this was related to the inability to meet rated power Finally to add insult to injury the power output was not stable with time varying by around 15W 3 of the rated power at full power It seemed likely that this was a thermal effect probably heating of the current shunt which could be expected to heat up since it was operated at high current in a confined space and being made of copper would have a considerable tempco of resistance This would explain the compression issue too Note a retest of the inverter one year later see Section 9 4 found that this power instability effect was no longer present Therefore there are grounds fo
35. Q1 2N3906 R1 3R3 200W SRO OON R15 100K 10 turn IRF530N x4 Ground R5 150R Fig B 1 Circuit diagram of dump load Fig B 2 Dump load interior view Size approx 200x250mm 209 Fig B 3 Dump load assembled 210 Appendix C The Smart Socket The proposed system needed some way of making electrical loads smart In other words giving them a means of telling the dispatching system how much power they required and how much they were prepared to pay and in turn having the dispatcher tell them whether it was acceptable to turn on The way of doing this would ideally be simple inexpensive and applicable to existing electrical appliances To achieve this backwards compatibility it would have to take the form of some kind of device that connected between the appliance and the power source sensed the power demand communicated with the dispatching system and optionally interrupted the power supply if the demand was not authorised It seemed that a logical place for such a device would be inside the power socket to which an appliance connected hence the name Smart Socket was coined The different kinds of power measuring systems were discussed at greater length in Section 6 3 2 To recap the simplest and cheapest option is a sensor that just detects the presence or absence of electrical current flowing through the load The wattage of the load is measured beforehand and is assumed always to be
36. SSUES 150 10 5 5 Concurrency and commitment 151 10 5 6 151 10 5 7 Combined heat and power dispatching esee 152 10 5 8 New Electricity Trading Arrangements eese 152 10 5 9 Politics economics subsidies eese 152 10 5 10 Strategic barriers to implementation eese entrent 153 10 6 AFINAL WORD sit ade eso eto tr e ER ER E E ERE E A EE E EN EEE 154 APPENDIX A BUILDING AN INVERTER esee eee ee eee ata tata asa to tanta totas 155 1 WHAT IS AN INVERTER 155 2 WHY BUILD AN INVERTER arenira itii a i R rete be aeter e EE CHE 155 A 3 e EE EE Ce E enero PE eee Re EE 156 4 rte e AIEEE stash RENE 157 A 5 BASIC DESIGN CHOICES saorte e e EG RR e te des er EE 157 AS Switching technologles te d e e e derent 158 ADD Circulltopolo giess sse v e t e e dieta e e oe su eet 158 A 9 3 Gontrolalgoritlms uer e e eee RN e eden 161 A 6 SIMULATIONS enti trt cr e RI C EEG EE NEU 163 Ad PRACTICAL DESIGN ISSUES isch etr EE RE RR RE RE REGARD 163 A 8 CONTROL CIRCUITS 168 4 0 1 PWM generator s re e HOT RE OP Er RM iH eren 168 4 0 2 Switching l09i6 o ee er Hn eee eet t ne ERE 168 A G 3 DIA AELE 17
37. The model would be combined with battery control logic and REDMan buying selling agents into a single battery management program named obviously BatMan Within this program collecting data on power bought and sold and its cost was easy enough the data being available from the REDMan subroutines The one problem was that REDMan s native units of energy are watt timesteps and the timestep nominally 5 seconds was quite poorly controlled in the prototype environment The solution to this was a conversion routine that measured each timestep by means of the computer s internal millisecond clock and so converted accurately to watt milliseconds and then trivially to watt hours The bought and sold values were accumulated in separate integrating registers multiplied by the respective prices and the difference taken 104 Profit pence Depth of discharge dimensionless Fig 7 1 Profit vs depth of discharge for a sample run of economic battery management 105 Collecting data depth of discharge was not too difficult either Data acquisition routines had previously been written which supplied the battery terminal voltage and current to any other program requiring it So it was a simple matter to write a coulomb meter program which ran once per timestep like all subprograms of BatMan multiplied the measured current by the measured length of the timestep and added it to an integrating regist
38. a dump load which was legally justified see chapter 6 because it was only activated during office hours when loading easily exceeded the inverter s output However it was not satisfactory because it made no use of the system s storage capabilities nor did it demonstrate economic dispatch of the inverter 7 2 3 Calculating inverter efficiency The initial approach to this was to measure it in rea time Since the DC input could be measured and the AC output was assumed accurately proportional to the control signal the losses and efficiency could be calculated There were a few problems with this however The ripple in the inverter current caused the measured DC values to fluctuate somewhat so the calculated loss figure was very unstable It could occasionally be seen to go negative so either the laws of physics were being broken or it was wrong Averaging it would have helped but at the cost of slowing the measurement down even further Even without averaging there was a 1 second delay between resetting the inverter power level and knowing the new DC input This was undesirable in algorithms that used a feedback controller and so a faster method was devised This meant compiling a look up table from laboratory tests on the inverter relating the control value AC output and DC input Knowing any one of these parameters the table could be used to find the other two immediately This of course assumes that the inverter performance is re
39. a little charge some more sit idle for a time and so on On the other hand most theories of battery functioning and ours are no exception assume a single discharge starting with the battery full and ending at the desired depth of discharge Adapting the model to this different regime of operation will require some more consideration of the details The general plan of action consists of allowing the battery to buy and sell energy as required all the time monitoring the depth of discharge and using the model to estimate the cost of wear and tear When this looks like exceeding the profit made by buying and selling energy something should be done about it that will cut losses That something is obviously to stop selling any energy and to allow the battery to recharge as quickly as possible until it is full Of course this process carries a cost too the energy for recharging costs money and there is still area susceptible to sulphation until the battery is completely recharged Both of these costs can be calculated if it is known at a given time how much energy is needed to refill the battery how much it will cost and how long it will take Knowing these quantities an estimate for sulphation during recharge can be added to the model TTR DOD t 2 Eq 7 8 and an estimate for the cost of recharging 102 DOD t co y nom nom Eq 7 9 where is the charging efficiency C is th
40. and see what happens Another connected problem is pricing The inverter has losses and has to pay for energy that it buys If it is to operate economically it has to make a profit on the energy it sells to cover the losses This means selling at a higher price than it bought One of the prices has to be known beforehand either fixed by human intervention or estimated by some other algorithm Then the other price can be calculated Once all these values are known there is enough information to present to the two dispatchers who will decide how much power is required of the inverter This will alter the losses so it is necessarily an iterative process which will hopefully converge Of course dispatchers which had knowledge of incremental cost could perform an optimisation by analytical means which would be more elegant and probably more accurate This will be discussed in greater detail in Chapter 10 7 2 2 Prototype The original inverter agent worked by requesting power from the DC side the buying price being set low and the amount equal to the maximum input the inverter 110 could handle Since Opportunistic flag was asserted the dispatcher allocated whatever surplus PV power was available The inverter was then controlled so that its DC input power was equal to this amount by measuring the input voltage and current and using an integral control algorithm See section 7 4 1 This was equivalent to using the inverter as
41. any DC component It happened that there were some surplus current and voltage transformers available so these were tried first of all D 2 1 Current transformer The CT was taken from a scrapped electronic wattmeter apparatus It was originally a 1 1 CT rated at 0 2A By connecting the original primary and secondary together in series and then adding a new 3 turn primary winding it was made into a 100 1 CT 218 capable of handling 20A It was used along with a 10 ohm burden resistor to give overall sensitivity of 0 1 V A D 2 2 Voltage transformer A small toroidal power transformer was used as the VT The toroidal type was chosen since it has less magnetising current leakage reactance DC resistance etc than the E I type and hence ought to reproduce the waveform with less distortion The transformer was chosen simply because it was the smallest toroid I could find It had a 240V primary and 15V secondary therefore a ratio of 16 1 The 15V RMS output would be too high for the soundcard input and some sort of attenuation would be needed D 2 3 Circuit The measuring circuit see Fig 13 1 was very simple Outputs from the VT and CT with burden resistor were taken to a pair of trimpots for calibration In the case of the VT a 100k resistor was added in series with the 10K pot to give an additional 10x attenuation Outputs from the pot wipers went directly to the soundcard line inputs current to the left channel and
42. appreciated that marginal cost facilities would be very desirable Therefore a revision of the dispatching algorithm to include it is recommended This might consist of adding extra information to the communications network a generator with 149 incremental cost would provide not one single price but coefficients modelling equation that described cost as a function of power level For simplicity it would be desirable to use a general equation applicable to all kinds of generator converter depending on the values of the coefficients However it would be possible to include a flag for selecting alternative models The dispatcher would set up the modelling equations for each generator including those with zero incremental cost which would be represented by all coefficients zero except for a constant term Equations representing the money earned by selling to loads would also be set up in the same way The complete system of equations would be solved in such a way as to maximise the money earned per unit of total generating cost 10 5 4 Multi domain issues The experimental system functioned properly with two domains however its dynamic behaviour was more sluggish than expected More thought needs to be given to the design of the software components that handle the connection between domains so that they can better decouple domains from one another There are a few possible ways of doing this For instance it might be
43. because transmission lines transformers etc have an inductive component The inductive reactance of the system is usually much larger than the resistance and so is responsible for most of the voltage drop Therefore connecting a capacitive load can reduce the voltage drop eliminate it completely or even raise the voltage above normal Capacitors may be permanently installed or switchable under control of an AVR system Another possibility is to use a synchronous motor since these have the ability to generate or absorb reactive power as required This may also perform mechanical work or it may be installed purely for power compensation in which case it is called a synchronous condenser The synchronous generators used in most large generating stations also have the same ability Recently power electronic technology allowed the creation of the static compensator STATCOM which performs the same function as the synchronous condenser but without moving parts 10 Modern STATCOM active filter systems are also capable of advanced functions such as reducing harmonic distortion 11 18 Of course reactive power compensation assumes that the network impedance is mostly reactive This may not always be the case particularly in the distribution network which uses lower voltages hence more resistive losses and can have a predominance of underground cables which are not inductive but capacitive Under these conditions reactive power compensat
44. comparator is U2A of Fig A 9 It compares the DAC output smoothed slightly by C2 and buffered by with the voltage across the current shunt amplified ten times by differential amplifier U1A and smoothed to eliminate high frequency interference by R5 R6 R7 and R8 set the amount of hysteresis while R9 is a pull up resistor required due to the design of the comparator chip open collector output The PWM waveform appears at U2A output and is fed to the power switching circuitry via a switching logic circuit described next A 8 2 Switching logic The main issue to be addressed is how to drive the switches in the proper sequence There are six separate switching elements and yet the PWM comparator only gives one output Some sort of additional logic is required to sequence the operation of the switches With reference to Fig A 10 the basic strategy of operation is this During one half cycle of the grid voltage the PWM waveform is applied to S1 S2 S1 receives the normal waveform and S2 receives its inverse so that S1 is ON when S2 is OFF and vice versa The result is an amplified copy of the PWM waveform at the junction of S1 and S2 While this is happening 53 and 54 are both OFF S6 is permanently ON connecting the other output terminal to ground During the next half cycle the roles must be swapped over S3 S4 receive the PWM signal while S1 S2 are off and S5 is ON 168 ag arri 1 doce Lzasd 1
45. connect to the grid the reference must be phase locked to the grid voltage Also the power output is controlled by varying the magnitude of this wave so to ensure accurate results the magnitude must be stable and controllable in an accurate and repeatable manner Various ways of generating this were investigated and there seemed to be two attractive methods The first was to use traditional analogue techniques The reference would be generated by a sinusoidal voltage controlled oscillator VCO with stable output amplitude A phase locked loop PLL would be used to synchronise this with the line voltage To vary the power level an analogue multiplier would be used to multiply the reference waveform with a DC power command voltage This would ultimately be derived from a digital to analogue converter DAC under control of a computer running dispatching software The main competitor was a microcontroller based system where all the processing is done by a computer program and a DAC outputs the finished reference waveform This system could be called direct digital synthesis DDS In this the reference waveform is stored digitally as a look up table LUT in memory At regular intervals successive cells are read from the LUT and sent to the DAC However for this application the basic DDS is elaborated somewhat To implement power 174 control each value from is simply multiplied by a power command value read in fro
46. during switching Transients of 10 20 volts were observed across only the inductance of a 80mm x 15mm x 2mm thick copper strip A complete solution of this problem would have required scrapping the existing power stage and redesigning it to a more compact layout which was not feasible due to time pressures Derating the unit to 550W proved to be an acceptable workaround The excessive THD was probably a result of too large a hysteresis band coupled with inadequate low pass filtering of the output So in conclusion the Mark Two inverter almost met the specification but not quite Considerable time and effort were devoted to improving the performance but in the end the above figures represented the best possible without a redesign There was no time to embark on a Mark Three and so the Mark Two unit was the only hope It should be borne in mind that the original goals represented the standard obtained by the very best commercial inverters Even though it did not meet the spec the Mark Two still outperformed many units currently on the market Also the only shortfall which actually affected standards compliance was the distortion This only required marginal improvement which could easily be done by adding an off the shelf EMI filter for instance Therefore it was still quite acceptable for experimental use 193 A 15 Protection Once the basic operation of the inverter had been proved it was time to consider a protection sys
47. end of every cycle reducing the inductance dramatically and causing spikes of magnetising current more conservative transformer design would be the obvious solution A 10 4 Latch up The final insult was that it occasionally suffered latch up This was a frustrating condition where once in a while all six MOSFETs would turn on simultaneously as soon as power was applied to the circuit The result was a complete short circuit across the DC bus normally followed by some kind of small explosion The root cause of this was that the MOSFET driver circuits were inverting i e a HIGH input to the drive transistor base turns the MOSFET off If the power to the driver circuits were to come up before the power to the logic circuits then the drivers would start operating while all their inputs were still low The solution would be to sequence the power supply rails so that the driver circuits were powered up last of all A 11 Mark Two The Mark One inverter had served its purpose as an experimental prototype However it was not powerful enough and in any case had design flaws which would require serious revision It seemed that the best course of action was to build another inverter With these shortcomings in mind work began on the Mark Two The most important goal for this was extra power 600 watts instead of the Mark One s 100 Achieving this extra power while meeting the 9096 efficiency target required some careful planning The fi
48. energy conversion scheme for connection to weak systems A S et al IEEE Trans Energy Conversion Vol 14 No 1 March 1999 36 Chapter 4 A new dispatching system proposed In the previous chapter the control systems currently used in power systems were reviewed as well as the likely problems when these systems are combined with and or applied to embedded generation Now the next step is to propose a possible new paradigm for power systems control The single grid control centre is replaced by large numbers of distributed dispatchers throughout the network Each of these acts as a broker through which software agents buy and sell energy on behalf of generators and consumers which they represent This will be referred to as distributed or embedded dispatching 4 1 Distributed dispatching The immediate question begged by this is What physical form and location would embedded dispatching systems take and how would they communicate with generation demand and each other This question is probably best answered by looking at the nature of the sub tasks involved in dispatching The computational and data requirements of these three components suggest different ways of embedding them 4 1 1 Embedding LFC and AVR As evidenced in 3 7 5 load frequency control is a task that requires quick action Therefore it is wise to place it where it has immediate access to the frequency measurement and the means of actuatin
49. for square wave is equal to the DC source voltage This is very easy to implement but a square wave will not satisfy the requirement for low distortion The total harmonic distortion of a square wave is around 55 The next step in complexity is to arrange for a period in the cycle when the output voltage is zero In the full bridge inverter circuit for example this is achieved by turning S1 S3 on hence shorting the AC terminals together The result is a square wave with pieces missing which can be arranged to have a peak to RMS ratio the same as a sine wave This is very useful when the inverter is used to power a collection of normally mains driven apparatus which includes some appliances functioning according to the RMS voltage and others requiring the peak voltage to be correct With an ordinary square wave the peak RMS ratio is always 1 so both conditions cannot be satisfied at once The harmonic distortion of this modified sine wave is also less than a square wave at around 25 but this is still too high to meet the specification Therefore it will be necessary to turn to more advanced methods The most popular and most efficient way of creating a genuine sine wave output is by pulse width modulation PWM This starts with a sinusoidal modulating wave at the desired output frequency and a triangular carrier wave at the desired switching frequency These two waveforms are fed to a comparator an electronic comparison cir
50. guaranteed latency and QoS Of course in the real world nothing is guaranteed Guaranteed latency just means that the latency is unlikely to exceed a certain value It does not mean that the system is immune to physical damage or malicious software or that data can magically be transported faster than an undersized physical network infrastructure will allow A system constructed along REDMan lines would therefore be vulnerable to breakdowns in the data network as well as breakdowns in the power network In all fairness it must be remembered that power companies already make considerable use of datacomms networks in scheduling their existing generation 4 70 5 7 6 Some sample agents To assist in testing two simple agent programs were written One was a test source which submitted a power and cost as typed in by the user The other was a test load which could have its price set by the user and which reported whether it was switched onor dumped by the dispatcher 5 8 Outcomes and conclusions This chapter was concerned with initial development of the dispatching algorithm The rules for dispatching developed earlier were transferred into a computer program and tested The relevance of optimal dispatching under the simplifying assumptions made in this system was investigated and it was found that the optimisation problem was trivial if not nonexistent This suggests that the economic dispatching could be improved by removing
51. heart of the dispatching engine valid Are the cost models built into the agents correct There is little hope of answering the first two which are subjective but the third and fourth will bear investigation 3 is a function of 1 and 2 It was known at the start of this project that economic dispatch or indeed any kind of coordinated dispatch of small scale embedded generators was a novel idea If 1 and 2 can be answered in the affirmative or even just 1 then something novel has been done 4 is obviously impossible to test by experiment with real generators Some sort of simulation could conceivably be made though which connected multiple REDMan instances together and evaluated the performance Also the experimental system contains two instances So to summarise the objectives of the experiments should be To test whether supply and demand are held equal both in the dispatching software environment and in the hardware To detect any errors in the supply demand matching either transient or steady state To validate invalidate assumptions made in dispatching algorithm To validate invalidate cost models of equipment To examine the behaviour of multiple interconnected REDMan instances Testing dispatcher This will entail two experiments The first is to scan the logged data and ensure that the dispatching engine s inputs and outputs all add up This will be quite simply achieved by importing the datalog files
52. if it were one source and one demand The number and size of domains in the network will be determined by other factors to be considered later Operations 1 2 and 3 are repeated at a regular interval the timestep When a source is ordered to supply power there is an implicit commitment for the period 44 of one timestep The same is also true for demands Means of monitoring to ensure that these commitments are fulfilled penalties for non enforcement and means of enforcing such penalties are not currently defined The timestep is uniform within a domain and all communications between sources demands and dispatcher are carried out synchronously with the timestep The system described by these rules is shown in simplified form in Fig 4 1 Only one domain is shown here a possible arrangement of multiple domains is shown in Fig 4 2 Renewable generation is included here on the premise that it can be dispatched to an extent e g constrained down See Section 3 7 6 45 Generation Fig 4 1 Rules expressed in diagram form 46 4 Fig 4 2 How multiple domains might be configured 47 4 4 2 Timestepping considerations The above processes will be repeated at most once per timestep One timestep must be at least the amount of time it takes for the information in one domain to be gathered the matching algorithm to run and the results to be redistributed There is
53. into a spreadsheet and summing the appropriate columns If supply and demand are truly matched then the sum of power supplied minus the sum of power demanded should always be zero It will also be important to examine the transient response of the dispatcher how long it takes to process sudden changes in supply demand and whether it settles cleanly to the new configuration or dithers around This can be done by extracting log file entries corresponding to times when changes occur and plotting them as a time series for 128 visual examination good presentation format would be as two stacked line charts one for supply and one for demand aligned in time Check agent programs assumptions The second is to measure the actual flows of power and check that they are the same as the desired flows within reasonable limits of accuracy This data may also be extracted from the log file but this is perhaps not the best way of doing things because these are the same figures used by the programs which control the power flow So they will appear correct even if the measuring transducers installed on the test rig are inaccurate It would be better to use a separate set of volt amp and watt meters There is also one power flow which is invisible to the system the battery s internal self discharge and losses This is not accounted for in any way by the dispatcher at present it is not clear what happens to it at all This should be inves
54. market for all of the power generated see 7 4 If the generator had been of the traditional type burning fuel that costs money with less 100 efficiency matters would be more complex For dispatching purposes the generator would be more like an inverter which also has to pay for its input energy and has an efficiency depending on the power level It might be possible to adapt the algorithms developed for inverters see 7 2 7 4 Dump load agent The task of the dump load is simple enough It only needs to get rid of surplus energy The definition of surplus energy is a different matter altogether In the early versions of the system the dump load was controlled by the generator agent to burn off the difference between the energy delivered from PV wind and the energy bought by the dispatcher This was not really in tune with the concept behind the system though A simpler and more elegant approach was to have the dump load controlled by an agent that bought any power available at a very low or zero cost By pricing the RE output at slightly below this figure Le very low zero or even negative then the RE generator would always be guaranteed a market for its total 114 output It should be borne in mind that in real life there are dump bads that do useful work such as off peak electric heating systems 7 4 1 Implementation Connection to the dispatcher was done using the standard communications routines use
55. more economical to make islands unstable and self destructive as is currently done but there seems to be no future in it 3 5 Impact of EG on existing control systems Adding EG to a network at the distribution level will change the power flows in ways that were never envisaged when the frequency control AVR and protection systems were set up Therefore these systems might not function as intended and so the agencies who plan power systems tend to treat EG units as a nuisance The problem here is real but fundamentally it is just because of a lack of co ordination between the EG units and the existing systems If that co ordination were present in other words if EG units were dispatchable they could probably have a positive contribution to the existing control systems 3 discusses this and proposes that EG and AVR should be handled together by a Distribution Management System This seems very similar to the system proposed in this work 3 6 Economic dispatching 3 6 1 Theory The goal of economic dispatching ED is to operate all the generators in a network together in such a way as to get the cheapest electricity overall This may be done 26 from a genuine profit motive or alternatively the money and profit involved may be notional and the real goal to minimise fuel consumption and hence environmental impact It will be discussed here as if real money and profit were involved but either way the concept is fundame
56. no longer in production A LabVIEW driver was written for it especially for this experiment C 3 Circuit description This description refers to Fig 12 1 The live line of the mains input LIVE IN connects to the current sensing diode network D3 D7 This is made from a bridge rectifier with an extra diode connected across the DC output terminals When a current passes the voltage developed across the diode network lights the LED in opto isolator 05 This is an AC opto isolator and so has two LEDs back to back converts the opto isolator photocurrent to a voltage which is buffered and Schmitt triggered by U1A The output of U1A operates a simple one shot circuit comprising D1 R2 and U2A The operation is as follows When the optoisolator lights UIA output goes LOW is discharged rapidly D1 and U2A output goes HIGH When the optoisolator goes off C1 recharges slowly via R2 causing a time delay before U2A output goes LOW again This time delay is chosen to be longer than one half cycle of mains 1 100 sec and so U2A output stays steadily HIGH even though the optoisolator goes out at every current zero The live output from the diode network goes to the load via relay RL1 Capacitor C2 allows a small current even when the relay is open so that the sensor will continue to 213 function Q1 Q2 are an ordinary Darlington driver which boosts the current capability of the control signal so that it can opera
57. of components This task could be off loaded by having the generator manipulate prices if it needed to continue generating at a higher level than the demand warranted it could attempt to increase demand by lowering the selling price This would be a job for artificial intelligence associated with the generator As argued earlier however control movement is often not a problem with the kind of small generators under consideration Lastly there has been no consideration of frequency control LFC at all This does not signal an intention to ignore it but rather to embed it at a lower level in the generator storage system apparatus itself for the sake of fast response LFC may override the requirements of economic dispatch on a temporary basis and in the longer term present itself to the economic dispatch system as a virtual generator or load which buys sells the amount of energy needed for frequency control This type of organisation is similar to the balancing market employed in some modern electricity trading systems These assumptions seem very gross and perhaps almost untenable But there are compelling reasons for simplifying matters to such an extent The system will have to deal with supply and demand data from millions of sources and react on the same time scale as changes in demand essentially in It will also have to be robust since its reliability will directly impact that of the electricity supply It was
58. possible to adapt the idea of tie line bias control from conventional AGC In this effort is made to hold tie line power flows constant except where the demands of load frequency control temporarily override this objective The setpoints for tie line power flows are calculated by economic dispatch In the system under consideration the inter domain connection would take the place of the tie line By arranging domains in a tree like hierarchy dispatchers at higher levels would see lower level domains as if each was a single generator load and by performing economic dispatch on them would be managing the tie line power flows There is no reason why this could not be repeated across multiple levels provided that the levels were decoupled adequately This issue seemed to be related to the variable timestepping system Proper rationalisation of timesteps used in the various domains and components would probably help to improve dynamics It is probable that the timestep should be the 150 same for all programs in a domain However timesteps may be made progressively slower as the hierarchy level became higher For example bottom level individual building subsystem within building dispatching might operate at a timestep of 0 1 to 1 second and top national level at a timestep of minutes to hours which is a comparable timescale to that of current national grid type dispatching The inter domain connection would need to perform explici
59. reaches of the distribution system the transformers have no taps at all or taps selected by hand in a set and forget style 3 3 3 AVR considerations for EG Most embedded generation plant presently does not participate in AVR This is for a few reasons Firstly individual embedded generators are normally too small to have any significant effect on network voltage Of course there are some large EG units and it could also be argued that a lot of small generators could have the same aggregate effect as a large unit However realising this would require some sort of management system to make all the generators work together This has been proposed see 3 5 below but never tested Secondly some kinds of EG plant use induction generators and so cannot perform reactive power compensation This may not be a great problem since reactive power compensation is less effective deep in 22 the distribution network anyway 21 On the other hand there are other EG types such as PV microturbines and variable speed wind turbines that incorporate a power electronic interface inverter This can simultaneously be used as a static compensator for AVR purposes 22 3 4 Protection Protection may be thought of as a control system too It monitors the operation of power system components such as lines transformers and generators and takes action if the component is in a potentially dangerous state That action is normally the opening of a c
60. shelf 625VA toroidal transformer with 230V primary and two 40V secondaries and 596 volt drop at full load regulation was chosen as a base First of all the magnetising current was measured Although the mains voltage was near 250V on the day of the experiment the current was very low less than 30mA RMS It would have been difficult to measure with more accuracy because the switch on surge would have destroyed a sensitive meter The transformer was also silent in operation with no buzzing In any case if the magnetising current had been higher the transformer could ha ve been modified by adding extra primary turns Next 10 turns of wire were placed on it and the voltage on this winding measured 4 99V Thus each turn gives 0 5V 182 But what voltage should the new winding be The inverter will malfunction if the DC terminal voltage ever drops below the peak AC voltage Now the DC voltage is nominally 24 But being supplied from a lead acid battery it could drop as low as 20 So the peak AC should be a little less than E 14V Leaving a little more J room for the inverter s internal 10 voltage drop design efficiency is 90 the result is around 12V therefore 24 turns Next calculate the thickness of secondary conductor required Assume that the 596 volt drop is shared equally between the primary and secondary so that the volt drop in the new secondary should be 2 5 It is also known that at full power the curr
61. so it may proceed if not it must remain off or resubmit at a different price If the amount of power required or the maximum price should change the submission should be repeated As a result of this it may turn out that the demand can no longer be fulfilled in this case it must switch off or resubmit at a different price This information will be supplied to a dispatching algorithm whose goal is to buy energy from sources and sell it to demands in an optimal manner or as close to optimal as possible The recommended algorithm for the time being is a simple rule based one embodying the three rules described earlier A more advanced algorithm may be substituted at a later date The price that the demand submits to the dispatcher in 2 is the price that it must pay irrespective of how much the dispatcher paid for the electricity This is of course the practical implementation of that cornerstone of capitalism an item is worth exactly what the market will bear It creates a deadlock that keeps things fair otherwise demands could ensure fulfilment by quoting enormous prices in the knowledge that their bluff would almost never be called The network may contain several such dispatching algorithms each dealing only with sources and demands local to itself The system of algorithm and local sources loads may be called a domain To interconnect different domains two connections will be used so that each domain appears to the other as
62. statement generates a retlw instruction for each data item note there is extra dummy data at the end org LUTBASE table addwf PCL F jump to the right place in the table dt 26 13 19 25 31 37 44 50 56 62 68 74 80 86 92 dt 98 103 109 115 120 126 131 136 142 147 152 157 162 167 171 dt 180 4185 189 193 5197 201 205 208 2127 215 219 222 225 2285 23T 3 dt 236 238 240 242 244 246 247 249 250 251 252 253 254 254 255 5 dt 255 255 255 254 254 253 252 251 250 249 247 246 244 242 240 8 dt 2236 4233 231 228 225 2222 219 44215 2125 4208 20555201 197 193 218 9 16772 L62 3157 a 25 36 115 109 9 dt 98 92 86 80 74 68 62 56 50 44 37 31 25 19 13 6 0 dt 128 1287 1287 7 128 128 1285 128 129 128 0128 128 128 1298 1298 128 8 e ck ck ck ck ck ck kk kk kk kk ck kk kx x RESET VECTOR e ck ck ck ck ck ck kk kk kk kk kk kx kx KKK org RESVEC interrupts off until we are all set up bcf INTCON GIE goto init e ck ck ck ck ck ck kk kk kk ck kk ck kk DO IT ALL HANDLER e ck ck ck ck ck ck kk kk kk kk ck KK KKK org INTVEC interrupts are stopped automatically when entering handler find out where the interrupt came from zero crossings get priority btfsc INTCON INTF if a z
63. that the specification is known the task is to design the inverter circuit that will meet it There are a number of different possible circuits but fundamentally all inverters work by using switches to periodically reverse a direct voltage So the two main design choices to be made are what sort of switches to use and what control algorithm to use for switching them on and off 157 A 5 1 Switching technologies Historically inverters have been made with every kind of switching apparatus such as rotating or vibrating mechanical contacts gas filled electronic valves and thyristors SCRs However in contemporary use the field is led by two special kinds of transistor The first kind is the MetaF On Semiconductor Field Effect Transistor MOSFET This device has a very rapid switching action and can be designed with a low resistance so that it will pass high currents efficiently provided that the voltage it has to stand in the OFF state is low MOSFETs designed to withstand high voltages have a much higher ON state resistance making them less efficient Whatever the voltage rating MOSFETs are electrically robust and difficult to destroy by excessive voltage or current Complementing the MOSFET is the Insulated Gate Bipolar Transistor IGBT When designed for high OFF state voltages this outperforms MOSFETs although the MOSFET is still best at lower voltages IGBTs switch rather slower than MOSFETs and
64. the air cycle efficiency which is a function of the compression ratio r Eq 2 2 State of the art engines currently have compression ratios of between 8 and 16 which give air cycle efficiencies of 56 and 67 assuming 1 4 for air respectively Again though in real life the best IC engines are about 40 efficient 4 when operated at their optimum speed and load In transport applications the widely varying speed and load reduce the overall efficiency even further The point of quoting all these statistics is to suggest that as regards improving the efficiency of heat engines it may be a case of diminishing returns If by some feat of materials science the absolute source temperature in a contemporary generating plant could be doubled without melting anything the Carnot efficiency would only increase by 29 which would probably mean at most a 16 increase in the actual efficiency That 1696 would probably be bought at an enormous cost in terms of elaborate materials advanced design and reduced reliability Another possibility might be to keep temperatures constant but make up the shortfall between the Second Law efficiency and reaklife efficiency by reducing irreversibility There are a number of innovative technologies such as combined cycle gas turbine plant and bottoming cycles for steam plant which are used to boost efficiency in this way The fundamental principle is to use several real cycles in order to approx
65. the assumption that incremental cost 1 Zero and re designing the algorithm appropriately A system was also created by which different programs within REDMan could communicate A networking strategy suitable for experimental purposes was developed by using existing Ethernet TCP IP and DataSocket components The network was built and the software routines required to use it were written Some speed tests of the dispatcher and communications were conducted and the results proved to be satisfactory Finally the dispatcher and communications module were combined into one program the REDMan Server The operation of this was tested by connecting it to mock source and demand agents which were very basic simulations of real world generators and loads In this way it could be verified that there were no bugs in the system At this point the core of the system was complete However its ability to control real life generators and loads with varying power demands and prices was still untried Therefore the next step was 0 try connecting it to some actual power equipment and this will be covered in the following chapter 71 5 9 References National Instruments LabVIEW http www ni com labview NeuraFbased Generation Control for Highly Varying and Uncertain Loads Shoureshi R A et al Proc 2001 IEEE Porto Power Tech Conf Decentralized Load Frequency based on Control Ishii T et al Electrical Engineering
66. the ripple problem The electrolytics still heated up but not excessively Fig A 16 shows the decoupling arrangement Later additional circuitry was added a controller for the cooling fan so that it only started when the inverter was running and an undervoltage trip for the DC bus The reason for using an undervoltage trip was that if the DC bus voltage fell too low the inverter would act as a rectifier instead due to body drain diodes in the MOSFETs and start to backfeed the DC side from the AC side This might cause unexpected behaviour of the circuit and possible damage The undervoltage trip made sure that the inverter would be completely shut down in a safe manner before the DC voltage could fall to a potentially dangerous level 186 Fig 15 Internal view of Mark Two inverter Overall size approx 250 x 300mm 187 Fig 17 Detail of power circuit showing transistors and busbars 188 Fig A 18 Mark Two inverter assembled 189 13 Testing the Mark 2 Once the unit seemed to be operating satisfactorily it was time to conduct some tests First and most important was the efficiency power control test For this the inverter was powered from a series pair of 12V lead acid batteries with facilities for measuring DC voltage and current draw The AC output was fed back to the local grid via a digital power meter made by Elcontrol Italy type VIPD The results were not terribly encouraging
67. the same This is sufficient for many types of load such as incandescent light bulbs electric heaters and the like More complicated sensors would measure the magnitude of the current flowing or better still measure both voltage and current and calculate the true power However for this first attempt a decision was made to use the simplest kind of sensor possible C 1 Design considerations It was decided to use current sensor that would give a binary output depending on whether current was flowing or not Current is usually sensed by having it flow through a resistor and measuring the voltage that appears across the resistor However this method posed a few problems in this application First was that the dynamic range of currents to be measured was quite large The design was to work with an ordinary 13 amp mains socket The largest load which this can be expected to carry is of course 13 amps The current sense resistor must be able to take this without excessive voltage drops and heating In this context excessive 211 heating is heat that cannot be dissipated within the plastic body of an ordinary UK wall socket i e more than about 5 10 watts and excessive voltage drop is anything more than about 2 to 3 volts Therefore the maximum current shunt resistance 1s set by the dissipation requirement at around 2 0 05 ohms The volt drop across this at 13 amps will be 0 57V only This is quite a sma
68. user is willing to pay for the energy taken from it to the cost of the damage being done to the battery by supplying that energy In order to calculate this ratio a means of costing the damage done to the battery is needed 7 1 11 A model for battery wear and tear Earlier it was seen that there are two main mechanisms of damage in a lead acid battery sulphation of the plates due to being left in a discharged condition and shedding of active material due to excessive depth of discharge It seems obvious that the main factors affecting sulphation are depth of discharge only discharged areas of the plate are affected and time the longer it is left the worse the effect Bearing this in mind a simple model for estimating sulphation can be proposed Consider an elemental area of plate a which is so small that it can only hold one quantum of charge So its depth of discharge DOD is either 0 full or 1 empty Next it is assumed that the probability of it becoming sulphated is zero if it is full and P if it is empty Once it has become sulphated it is out of service forever So an expression can be written for the probability S of this small area becoming sulphated in a given time interval S P t DOD Eq 7 3 From this can be derived the number of small areas Ns which are liable to become sulphated in this time interval This of course assumes that the sulphation of an elemental area is a random event and that the total number of
69. very general brief To reduce the harmful effects of fossil fuel combustion At the outset this was easily mistaken for a technological problem and it was tempting to believe that there was some kind of magic bullet to be discovered which would make engines and boilers more efficient or which would make renewable energy sources so cheap reliable and tempting that it would not even be worth bothering with oil The first task undertaken was a literature review trawling books journals and the Internet for background information This gave cause for suspicion that the problem was of a different nature There were many examples of new technologies great improvements over the status quo in terms of combustion efficiency or harvesting of renewable energies like biodiesel the Coates engine cogeneration affordable electric drives for cars and even the old fashioned bicycle The magic bullets were all ready and waiting Why was nobody pulling the trigger Of course energy use is not a technological problem at all but a complex socio political one Reactionariness and fear of change are embedded into every society and institution denial is a part of everyday life and people will probably never drive electric cars because they don t make a revving noise Large amounts of money are invested in maintenance of the status quo short term profit is pursued irrespective of long term consequences and there is nothing that engineering can do about a
70. voltage to the right D 3 Software The software was written in NI LabVIEW using the sound input capabilities and signal analysis functions D 3 1 Data gathering The sound input was set up to provide a stream of stereo 16 bit data at 11025 samples per second This is one of the standard sample rates used by computer sound hardware others are 22050 and 44100 second The sample data piled up in a buffer and each time the buffer became full the chunk of data was passed to the analysis section of the program There was a tradeoff here As the chunks became larger the analysis would be more accurate being the average of a greater number of mains voltage cycles But each chunk would take longer to accumulate and hence the display would be slower in updating The buffer size was made adjustable in 1K 2 9 increments to help in exploring this tradeoff A 16K buffer hence an update period of 16384 11025 4 20 37 sec was found to give good results D 3 2 Analysis The buffer contents were converted into two arrays of 16 bit 2 s complement integers one for the left current channel and the other for the right voltage channel Next these were changed into double precision floating point format and multiplied by appropriate calibration constants The result was one time series 1 D array of voltage samples and another of current samples These were processed in three ways First corresponding elements of the two arrays we
71. was beyond the scope of this work 228 Electrical results output from libe Graphical output Period Tue 1 Jun 00h56 to Tue 1 Jun 823157 Year31999 Focus gt gt Lines cables Time steps sim 6m output oo nnn Con 8 Loss phi Con 8 Trn phi Power flow b Transmitted power c Power loss Power factor p f d Transmission p f Clear all selections Set axis scale Add another profile Change to tab stats Hide show output Draw graph 1 00 1 00 3 00 0 3 6 3 12 15 18 21 24 Time Hrs on 8 Trn ph kM Help Exit Fig E 1 Results analysis from ESP r showing inverter power output and losses 229 E 6 Conclusions This project provided a rudimentary model of an electrical inverter for the ESP r simulation environment It also threw up a lot of difficulties with using ESP r for the purpose envisaged It became obvious that a great deal more development work would be required before ESP r would be capable of modelling the kind of renewable energy systems in question In particular control algorithms for electrical network components would need to be added and to judge by the amount of effort required to implement the inverter model it seemed that this would be an extremely long and difficult task Therefore this line of work was abandoned A more fruitful alternative mig
72. with but as a first approximation it could be assumed that the amount of material shed in the course of a given discharge is proportional to the depth of the discharge Then by a similar line of argument to the previous section P C max DOD 1 CTF Eq 7 7 where the various factors are again lumped into an empirically derived constant CTF the number of cycles to 10096 D O D required to destroy the battery max denotes the largest value observed over the cycle Of course this model assumes that the proportionality is linear in other words if discharging the battery to 50 dislodges one unit of plate material then discharging it to 100 will dislodge two units and so on Proving this experimentally is such a long and involved process and 101 the result of such limited application that it is hardly if ever done and so it proved impossible to find data which would confirm or deny this assumption Actually performing the experiments is beyond the scope of this research So this model should not be considered accurate but rather as a temporary measure which is just good enough to help prove the general concept 7 1 12 Practical implementation In order to implement the model thought needs to be given to the conditions under which the equations are valid A battery in energy storage service with REDMan will probably have energy flowing in and out almost at random So the battery may charge a little discharge
73. with LabVIEW which includes printed source code is in black and white Unfortunately LabVIEW has no alternative way of representing its VIs The structure and appearance of the block diagram and front panel are saved together in a binary vi file of a proprietary format This file is essentially the source code and the only way of viewing editing compiling running it is through the LabVIEW GUI itself Some other graphical languages do translate the graphical commands into human readable text based code often or similar as an intermediate step in the compilation G 10 References For further information the on line help within LabVIEW itself is probably the most valuable source There is also a lot of information available at the National Instruments website http www ni com labview 245 Appendix LabVIEW programs This appendix provides information about the LabVIEW programs written in this work of the programs may be downloaded from the ESRU website at http www esru strath ac uk H 1 testbed software A description of the individual programs that drove the experimental testbed Bm2 vi BatMan battery manager version 2 Full listing of bm2 H 2 1 Buys and sells power to control and SOC estimator subVI battery state of charge 2 Daq2 vi Data acquisition version 2 Operates DAQ hardware and makes results available to all other programs through global variables Dumpcon vi
74. 0 1 units Note that these numbers are of no special significance they are chosen at random Now work through the algorithm specified in the flowchart Sort sources in order of price Take first demand from source list All of this will be supplied by the first source and the cost will be 0 17 which is less than the 10 that the demand is prepared to pay so the transaction can go ahead The new source list Next demand is 9000W If this was to be supplied it would use up the 8800W remaining from the first source the 72 of the second source and 9000 8872 128W of the third The cost would therefore be 8800 0 17 72 1 2 128 3 9000 0 218 This is less than the 1 that the demand will pay so again this transaction will be allowed The new source list Now for the final demand The remaining power has a cost of 3 but the demand is only prepared to pay 1 Therefore it will not be allowed and the dispatching operation ends here Next step is to feed the same source and demand lists into the dispatching program Fig 5 2 following page shows the output of the program when fed with this data input in fields marked 1 2 output in fields 3 4 and it can be seen to be identical to the output predicted by hand calculation This is a simple example in the course of debugging more complex test sequences were used to verify absence of rounding errors proper operation when demand exceeds supply etc 58 Jomod
75. 108 Conversion Accumulator REDMan to kWh end REDMan load generator Buying selling control Conversion to kWh Sampling Reset State integrators and kWh machine conversion Battery Trip points voltage ETE Peukert s Wear and tear niegrator Rule estimator current compensation Integrator Profit calculation From DAQ Peak detector Fig 7 3 Functional block diagram of the battery management program 109 7 2 Inverter agent 7 2 1 Design To a first approximation this would appear to be a simple piece of programming The inverter agent is a double headed combination of the ordinary load and generation agents One head connects to the dispatcher dealing with the DC network where it appears as a load The other head connects to the AC dispatcher where it appears as generation But complications soon arise How much AC power should the inverter advertise for sale As much as it can get from the DC side less its operating losses But what if not all of the advertised AC power is bought The answer is that the inverter should be governed by the available DC power or the bought AC power whichever is less Unfortunately this is not as easy to implement as it sounds The nature of the REDMan system is that it cannot tell you how much power you could have You either get the amount you asked for or less than you asked for or none The only way to find out if there is surplus power is to try and buy more
76. 2 A 8 4 Reference generator isses esee tette tenerent 174 4 6 5 CHOOSING a DTOGeSSOT d use te nr RE RE E ae OT Met HERES 175 derer e eH Hee e dee n Rr e EH REA 176 Suppott CUGCHHTys oet RR HO 176 A 9 BUILDING AND TESTING THE MARK ONE eese eene nnne nennen trennen seen 177 10 LESSONS LEARNT FROM MARK 180 AJOI Odd apike CERE 180 A 102 Excessive e ERR i be HER E a e Y 180 A 10 3 TOO m ch distortion sen ie e eire eret et aetate ke i 180 A 10 4 Latch up iius eate deae eite tee aede che iiie a aide e RS 181 MARK TWO leet eve acta see eu 181 ATIA Loss budget tace etate ii acie acid 181 LERNTE 182 A d 1 3 Current sense resisto a s ete te RE RENTRER 182 SLL AA 182 Al 1 5 Transformer REO OO cette 182 6 Filter components 184 12 ASSEMBLY AND SNAGGINQG 185 A 13 TESTING THE MARK RR Sa Ee RE dee eds 190 14 LESSONS LEARNT FROM THE MARK 2 esee nennen 193 A 15 PROTECTION oett taie RI e ERES RR EGO ECHO T RN Ue ERE UE 194 A 16 NOTES ON FIRM W
77. 2419221 0 p MOJJE 1 151 ur qoga 20 NO 1 1431 p 151 224nos ur s3ounos woaz 34104 mod Jo jo 181 c SOOINOS 30 15 00 000 000 000 ozi L zro poooze 14 G Durqojeur Fig 5 2 User interface of dispatching algorithm See text 59 5 4 2 Speed test Another matter of interest is the speed at which the program operates and the computing resources used for its operation Of course as discussed in Chapter 4 the eventual application will probably be some sort of embedded microcontroller whether built into a generator inverter etc or in a stand alone dispatching box of some kind However for the purposes of this initial experiment it is important to make sure that it will execute in a reasonable time on an ordinary PC running LabVIEW By reasonable time is meant that it should complete within the proposed timestep of one second and still leave a generous allocation of time for communications routines and other components of REDMan running on the same machine The results also give some idea of the embedded computing resources that might be required in future and so are presented here Unfortunately measuring the speed of the algorithm is not totally straightforward because it is not constant Due to conditional statements within the algorithm the time taken to execute will depend in a complex way on the values of data
78. 270 89 2562 5 1 81818 55 140 302 5 24 15 13 85 334 4775 298 89 09418 4 5 14876 60 153 330 24 07 15 4 370 678 328 88 4865 2 0 60606 65 165 357 5 24 16 6 398 4 353 88 60442 4 5 125874 70 178 385 23 97 18 25 437 4525 381 87 09517 4 1 03896 75 191 412 5 2394 19 5 466 83 407 87 18377 5 5 1 33333 80 204 440 23 85 20 9 498 465 435 87 26791 5 1 13636 85 216 467 5 2385 22 4 53424 457 85 54208 10 5 2 24599 90 229 495 23 75 23 9 567 625 482 84 91522 213 2 62626 95 242 522 5 2372 25 6 607 232 508 83 65831 14 5 2 77512 100 255 550 23 6 27 1 639 56 532 83 18219 18 327273 Table 9 4 Inverter re test data 139 9 5 Transient behaviour An investigation of transient response was made by scanning the logfiles for areas where rapid changes in power level appeared to take place and then examining these in further detail for any evidence of unexpected behaviour 9 5 1 Typical dispatching behaviour A 9 minute period characterised by rapidly varying PV power was found which demonstrates the function of the dispatching and inverter control at this time Powers vs time are graphed in Fig 9 2 PV generation inverter input inverter output and dump load Results can be interpreted as follows At first the inverter input tracks the PV output This is in accordance with the pricings set which state that the AC load can only afford PV power At sample 15 the PV generation becomes enough to supply the 50 watt AC demand through the
79. 9 is compiled from several sources including textbooks websites books and magazines aimed at renewable energy and electric vehicle designers e g 3 4 5 6 96 7 1 8 Estimating capacity The traditional method of finding the state of charge is by measuring the specific gravity density of the electrolyte Since the acid concentration is proportional to the state of charge and the densities of water and acid are different it is easy to infer the state of charge from the density of the mixture This method is accurate provided that the water and acid are evenly mixed throughout the cell the temperature dependence of density is taken into account and any water lost from the cell is replaced before measurement The measurement is done using a hydrometer which is inserted into the cell and sucks up a sample of electrolyte into chamber containing a graduated float Better quality hydrometers also include a thermometer for temperature compensation The hydrometer method has been used for many years and is simple and inexpensive the equipment only costs a few pounds However there are a few drawbacks it is difficult to automate because automatic sensors of density are complex and expensive and it is not applicable to sealed battery types e g gel cells where the electrolyte is not accessible A readout of the estimated remaining capacity or run time allows planning of recharges and so makes battery powered equipment more
80. ARES 195 AUT SPEOWCHART e e EUER dbi de tete 197 AAS e Oe 198 19 REFERENCES EEEE EE S 204 APPENDIX B A VOLTAGE CONTROLLED DUMP LOAD 206 1 DESIGN CONSIDERATIONS estet ates terret E EXC e Ro Et anced EAE 206 xii 2 CIRGUIE itt ho cim ite eie 207 B 3 COMPUTER INTERFACE p Rente gerit erre ee 208 BA CONSTRUCTION wees Atc nete e e tede ge eed Re tet ree cde 208 APPENDIX C THE SMART SOCKET eese esee eee eeeseste sensn tenens ense tenen ensssesenes sensn enesseseeesesssss Z1 L C 1 DESIGN CONSIDERATIONS retener Ee Ee Feste 211 2 INTERFACING 213 CIRCUIT DESCRIPTION FERE Fee REY ERE EE FRE RE EYE RECTE 213 APPENDIX D POWER MEASUREMENT 217 D 1 DESIGN CONSIDERATIONS eere eene 218 D 2 HARDWARE eere isc ae RG ase as Se ien idi RB 218 Current RE 218 1 2 2 Voltage transformi ao i ettet me Re PU I RRHIR GE ERU S 219 D 2 3 CIVCU i oe nt ete irum ee 219 D 3 SOELWARE D 3 1 Data gathering D 3 2 Analysis paseo ede n o hn ted a ee o ee Y ees DA CALIBRATION AND TESTING
81. AX D 256 MAXLEN LAT D 10 the absolute max timer setting define TREF D 128 MAXLEN 2 the timer reference point to aim for define THI TMR F TOL upper and lower tolerance bands define LO TMR FTOL define TARGET SPC 1 the target step index value defineCTR D 255 number of cycles to elapse before run define CTS D 10 number of frequency violation cycles before shutdown Protection states define SYNCING1 in process of locking to grid defineREADY 2 locked and ready to start power stage define RUNNING 3 power stage started define TRIPPED 4 tripped out defineASLEEP 5 shut off due to being idle for a while Working registers power equ Ox0c Throttle setting output power control index equ 0 0 current index into lookup table H byte equ 0x0e multiplier working regs 198 L byte equ Ox0f 7 as above mulplr equ 0x10 working reg for multiplier count equ 0 11 multiplier loop count timer equ 0x12 what the timer value should be based on errc equ 0x13 phase lock coarse error ETHE equ 0x14 phase lock fine error tmrtmp equ 0x15 timer saved value wtemp equ 0x16 stemp equ 0x17 run equ 0x18 ready to go when run 0 pstat equ 0x19 protection status fvi equ Ox1A frequency violation counter equ Ox1B Cycle counter e ck ck ck ck ck ck kk kk kk kk kk kx kx KKK LOOKUP TABLE 7 contains a half cycle of rectified sine wave note the dt
82. Agent representing the dumpload in the DC market Receives data on dumpload current from daq2 Grid2 vi Agent representing the power grid in the AC market Invman7 vi Smart inverter management version 7 Listing in Section H 2 Buys power from DC market and sells to AC market Load2 vi Agent representing a load Used for testing Matchingl5 vi REDMan dispatching algorithm Listing in Section H 2 246 Regen vi Agent representing RE generators PV wind turbines in the DC market Calculates available power using data from daq2 Server3 vi REDMan dispatching server version 3 Handles communication between agent programs and dispatcher matching15 for the DC market Copy of Same as server3 but for the AC server3 vi market Socket2 vi Smart socket software version 2 Agent representing loads connected to the smart socket strip Statistics vi Calculates hourly daily and weekly RE generation statistics Receives data from daq2 247 2 Block diagrams source of selected Vis H 2 1 BatMan battery management bm2 vi bm2 vi Connector Pane Front Panel 248 Block Diagram Server URL abe BatteySOC Target profit p cycle EDWearCost 2 Equalise Voltage exceeded Equalise period elapsed Po Float voltage exceeded a Demand for power Economic dispatch warming 249 Hidden parts of block diagram I Ini
83. DC wiring were negligible it was important to use the shortest and heaviest cables practical and not to include more fuses and shunts than absolutely necessary 84 E u oe 10 ui 10 Current shunt 100 Amp 30 Amp Safety disconnect Dump 24 T Voltage divider Fig 6 2 Wiring diagram of DC equipment terminal board 85 Fig 6 4 Ducted wind turbine 86 l board and battery bank Fig 6 5 DC termina 87 8 8 8 Aa 6 6 4 4 PTEMA S 8 8 820 625 Ue w amp ws e em i Pama rai Fig 6 6 Complete experimental set up 88 On a similar line of reasoning the connections for voltage measurement were made directly to the terminals of the battery thus establishing it as the theoretical node of the DC circuit In other words the powers measured would correspond to the powers entering or leaving the battery This choice was deliberate because battery management was expected to be the most demanding application in terms of power flow measurement accuracy With respect to the AC side the inverter was plugged into an ordinary wall outlet mixing its output with the existing grid electricity This may seem dangerous since power would be present on the pins of the inverter s plug if it was pulled out However in practice it is acceptable due to the inverter s fast acting loss of mains protection Pullin
84. DISTRIBUTED DISPATCHING FOR EMBEDDED GENERATION Stephen J Conner BEng A thesis submitted for the degree of Doctor of Philosophy Energy Systems Research Unit Department of Mechanical Engineering University of Strathclyde Glasgow United Kingdom August 2003 2003 Stephen J Conner The copyright of this thesis belongs to the author under the terms of the United Kingdom Copyright Act as qualified by University of Strathclyde Regulation 3 51 Due acknowledgement must always be made of the use of any material contained in or derived from this thesis ii Imagination is more important than knowledge Albert Einstein ACKNOWLEDGMENTS First of all I would like to thank my supervisor Professor Joe Clarke for advice guidance and proofreading and corrections of the piles of paper I dumped on him I am also grateful to the Carnegie Trust for the Universities of Scotland who funded this project Thanks are also due to my fellow students researchers and academics in ESRU who were always full of helpful advice and interesting ideas In particular Nick Kelly and Iain Macdonald who gave me a lot of assistance in the computer simulation side of matters and Paul Strachan who introduced me to LabVIEW an amazing piece of software which I used almost everywhere This project involved a lot of lab work and a lot of custom built equipment I am very grateful to all of the technical staff who made things for me
85. P benefits from smaller scale since the heat rejection temperature of each individual generator can then be tuned to suit the process requirements e g 50 9C for hot water 150 9C for industrial steam And if these small generators are connected with the existing electrical network there could be a reliability benefit due to the redundancy of multiple power sources This is a contentious point of view though which will be examined at greater length in a later section Distributed generation can also help exploitation of renewable energy RE Important forms of RE such as wind and solar power are diffuse by nature They can only be captured in large quantities by using a large number of distributed small generators covering a large area It may be possible to embed large numbers of RE generators in on existing buildings and structures Finally electrical transmission losses are greatly reduced since the electricity is not transmitted over any appreciable distance For nonrenewable sources though this advantage is perhaps not as great as it seems since some other fuel must be transmitted instead Transporting the fuel to the generator location uses energy too and this can be thought of as a transmission loss 2 41 The technology of EG Really EG is just a return to Edison s powerhouse on every block system which was abandoned 100 years ago It may well become attractive again because of improvements in small generato
86. RATING PLANT ce ERRARE RE EE EO ETE E EET RENS 75 6 2 1 Dispatching PV wind energy eee 75 6 2 2 Using batteries x i 76 0 2 3 InVeftersa se e idet at 77 6 3 CONSIDERATIONS IN SETTING UP T HE 5 TI 6 3 1 Choosing voltages and power levels eese eene tete 79 0 3 2 Measuring power TlOW a e ede Ret ste dede deest 80 6 4 SETUP AND COMMISSIONING csccsssssssssessssesssscscescssescesesscssssessescsscsssecssescnscassecsecsesecaeeensscaesecseseesecaes 81 OA T Equipment listet ate s t aded eel 81 edid 82 0 43 84 6 44 POW EP WIFIR dee e RO een e 84 6 4 5 Stenal WIR ee e ERROR esas IG RR RP ERU AR ER EMEN ER Nt e OR Need 89 64 6 Testing soie eee vd 90 6 5 5 90 6 6 REFERENCES Rm 91 CHAPTER 7 AGENT SOFTWARE tona en saos tassa ene ns to tenen sae es ta tonne sense ton saeua 92 7 1 BATTERY MANAGEMENT AGENT eese eene sensisset seas ea sensn 92 TN IM BETTE 92 7 1 2
87. RMWARE VERSION 3 03 WITH PROTECTION E TE HE He e e A Ae e Ae kk KEKE e e AE Ae Ae He e e STRICTLY C 2000 1 STEPHEN J CONNER ESQ BENG SO HANDS OFF E e e GN e e e ROMS REISK UN Nene HI RES VERSION INTENDED FOR PIC16 C F 84 10 WITH 9 8304 MHz CRYSTAL PEAR KIC e GR E ASSEMBLER SETUP PRERRERKEERERER EERE E RAR What kind of processor are we using LIST P 16F84 include file gives names to special function registers include P16F84 INC processor config flags watchdog on code protection off xtal oscillator etc CONFIG WDT ON amp HS OSC amp PWRTE ON IDLOCS H 1234 code origins define LUTBASE 0x300 lookup table base address define RESVEC 0x00 j power on reset vector define NTVEC 0x04 vector for one and only interrupt define LUTPCH LUTBASE D 256 high bits to poke to PC Magic numbers define FCY D 100 mains frequency Hz defineFTOL PASE mains freq tolerance in 128ths of a cycle define FOSC D 9830400 crystal frequency Hz define SPC D 128 wave steps per cycle defineCPS FOSC 4 FCY SPC clocks per wave step define LAT DITE total timer interrupt latency define D 104 maximum number of instructions in a step define ISAT SPC D 14 index saturation value define TMR D 256 CPS LAT initial preload for timer allowing for latencies define TM
88. T 4T CONCEUSIONS EA CHAPTER 5 DEVELOPING THE REDMAN SOFTWARE cccsssssssssscsseasesnsenseasssnsensesssensennsensenees 52 Dele THE DISSATE ER zoe ee neon oe t eee odit etie miis 52 5 2 DETAILS OE PROGRAMMINQG eec eii ceris 55 5 37 WRITING THE PROGRAM aanita a ee tiae eq eee 55 56 9 0 1 Ethernet networking E 63 2 06 23 DS TE DataSOCk6t s dtt ed eie eei e ee ee ede te ees 64 2 6 3 Testing DSTP DaiasSQCket o e ost eet 65 9 6 ReSUlisa 66 57 REDMAN COMMS ROUTINES e e t ede erre eee 67 2 7 T Dat items to be transferred eee ede teet 67 2 72 Allocatng index et edet te Dt ees 68 D5 Feds Timing and synchronisation nane e ette te e ete e ee e Ses 69 5 7 4 Realisation Of the system estet tete e tenente nete terere nnne trennen tenera 69 2 5 Laiency quality of service ete eae ee ee ie e nee eee eee e een 70 3 756 SOME sample agents e 71 5 8 OUTCOMES AND CONCLUSIONSG cccccssssssessssessesessessssessesesscsesscssescsscsseecsscessecassecsecassecsececnecsesecaeseesesaes 7 5 9 REFERENGES no Re TRO RR RR DE 72 CHAPTER 6 BUILDING A TESTBED tenen enta tones ta tan to tones ases tasas snos ense ton en 73 6 15 GRID CONNEGTED WORKING t ERR AOT ED E EE RC RR RE 7A 6 2 GENE
89. TUS C if errf index incf timer F add one to timer value movf timer W movwf TMRO now load the timer load up some registers movlw 0x08 set loop count for multiplier movwf count movf power W transfer current power setting to movwf mulplr temp register multiplier will mangle it fetch the right lookup table entry 202 movlw LUTPCH load PC high bits latch movwf PCLATH because the table is in a different page movlw LUTBASE put base address in W addwf index W add current wave index to W call table here goes nothing we return from lookup table with wave step value in W multiply by throttle value using microchip example multiplier code mul8x8 Multiplier is in mulplr multiplicand is in W Answer comes out in H byte the lower 8 bits are ignored It takes 73 cycles clrf H byte clear working regs from last time clrf L byte bef STATUS C Clear the carry bit in the status Reg mloop rrf mulplr F Rotate the multiplier right through carry bit btfsc STATUS C Test the carry bit if not zero addwf byte then add W to the high byte ert H_byte F Rotate high byte right with carry from addwf rrf L byte F rotate low byte right through carry decfsz count F decrement count and test if not zero goto mloop then loop again Test to make sure we are in run mode movf pstat W sublw RUNNING btfss STATUS Z goto clr Send out to D A LSB will be ignor
90. The pin is briefly reconfigured as an input at every zero crossing to sample the detector Scaling is done by a software multiplying routine because the PIC used has no hardware multiplier This uses well known arithmetic techniques to multiply two unsigned 8 bit numbers together taking about 77 instruction cycles to do so which leaves about 115 instructions per conversion for everything else The least significant 9 bits of the result are discarded in this application 196 A 17 Flowchart Initialise onboard peripherals Wait for interrupt Timer What was interrupt source Restart timer Zero crossing More than 128 Slow timer down Less than 128 timer up How many cycles since last zero crossing 128 Leave timer alone No Is timer rate within tolerance Increment violation counter Decrement violation counter Is counter gt threshold bus voltage OK 197 Read sine Multiply result by power command Increment lookup table pointer Some operations omitted simplified for clarity Enable power circuit No Get Is power power command zero Yes Disable power circuit A 18 Listing This program may also be downloaded from the ESRU website at http www esru strath ac uk p RW Woo upset e o eue oo s pue seed eoe as ap AE ous GRID INTERTIED INVERTER FI
91. This removed the rail to rail requirement so allowing a wider choice of op amps Second was a transistor switch allowing the microcontroller to disable all the MOSFET drivers at once so curing the latch up problem and allowing the inverter to be totally shut down Third was a great enlargement of the power circuit with 24 MOSFETs instead of the original six In fact the power circuitry could not be put directly on the PCB because the copper was not thick enough to carry the current So the board was used only for mechanical support and the current was carried by brass and copper bus strips joined together by nuts and bolts Considerable thought was put towards laying out the power circuitry to give the shortest current paths and smallest current loops possible and hence minimum resistance and inductance The drain leads on the MOSFETs were not used contact being made through the metal tab of the package instead This gave a lower impedance electrical contact and also better thermal contact The latter was not of great significance though because the transistors were not expected to dissipate very much power in this high efficiency design In fact there were no actual heatsinks as such the brass connecting strips that carried the transistors were just made a little larger and thicker than electrically necessary to help in carrying the heat away See Fig A 17 This design decision proved to be reasonable with transistor case temperat
92. ach client One complication in this scheme is that a single DataSocket URL represents a connection that passes data in one direction whereas data flows both to and from a client The solution was to define not one but two URLs for each client one for each direction of transfer The following naming scheme was adopted for the URLs Dstp hostname redman client server load gen number Where hostname identifies the machine running the DataSocket server application lt clientlserver gt is client for the connection that sends data towards the dispatching algorithm and server for the connection that sends it towards the source demand agent loadlgen is load if the agent is a demand agent and gen if it is a source agent and number is an index number uniquely identifying each agent For instance the connection between the first generator agent and the dispatching algorithm would be specified by the following two URLs Dstp localhost redman client gen 0 Dstp localhost redman server gen 0 This scheme is fairly complex and with a reasonable number of agents in use would generate large numbers of similar URLs This proved to be a nuisance during early development of the system where a good deal of errors were made simply through 67 confusing or mistyping URLs In order to make the process less prone to error program was created to build the URLs automatically a
93. added on top of the payload itself By using a packet sniffer program the traffic between the DataSocket server and the client program was intercepted Each transfer of a double precision floating point number was observed to use one packet of approximately 600 bytes in size It is hard to be more precise because the size of the packet depends on the point in the protocol stack where it is intercepted Given that an IEEE format double float uses only 64 bits 8 bytes this is somewhat disappointing although not very surprising given the amount of essential routing information that goes into a TCP IP packet besides the payload Related to this and also of interest is the load imposed on the computer s CPU by the DataSocket routines If this was excessive it might not leave enough CPU time for the proper functioning of other REDMan modules To test this the CPU load was measured using Microsoft Windows System Monitor application while using the previous test program to write and read fifty DataSockets per second The result was a load of almost 40 on a machine with 333 MHz AMD K6 2 processor 5 6 4 Results Performance of the dispatching algorithm and communications system has been measured The dispatching algorithm runtime was found to vary roughly as the cube of the problem size whereas the communications runtime varied linearly as the problem size With the full processing power of a computer similar to the 300 and 333MHz test mach
94. al implementation 6 3 1 Choosing voltages and power levels The choice of AC voltage is of course predetermined In the matter of DC voltage there is some choice but it is limited by the availability of components There are four PV modules two each of different kinds and all nominally 12 volts For good performance PV panels that are connected in series should be matched Therefore it is only possible to connect two in series and so the maximum possible voltage will be 79 24 And while lead acid batteries are available in other voltages the most common type are also 12 volts Therefore the choice is between 12 or 24 volts Since the higher voltage will allow more efficient operation with less PR losses in the wiring and inverter circuitry 24 is the logical choice 6 3 2 Measuring power flow As far as DC is concerned this is simple Provided that both voltage and current do not have a significant AC component the power is the product of mean voltage and mean current Therefore the power balance can be measured by measuring the DC bus voltage and each current entering leaving it The powers calculated from this will be those entering leaving the node where the voltage is measured These will not be quite the same as the powers generated or consumed because of power losses in the connecting cables The impact of this error can be minimised by designing for low voltage drops and measuring the voltage at a carefully chosen ce
95. ammer However while it had an ample 1 024 words of memory its maximum speed was only 2 5 MIPS This would be fast enough though if the precision were reduced to 7 bit and 64 points used per quadrant instead of the 128 implied by the precision This would still provide a good enough waveform to meet the distortion spec A 8 6 Firmware The firmware program is listed at the end of this Appendix and an explanation of the code is also given A 8 7 Support circuitry The PIC does not quite do everything by itself It requires a few supporting components the circuit is shown in Fig A 12 The quartz crystal is a standard part the frequency of 9 8304 MHz may seem odd in fact it was a convenient multiple of 50 Hz 50 cycles per second 2 half cycles per cycle 128 D A conversions per half cycle 192 instructions per conversion 4 clocks per instruction 29 830 400 176 Also needed is a digita to analogue converter which is formed by resistor network R5 R20 a classical R 2R ladder R21 R22 allow the output voltage to be adjusted QI and U5 multiplex the 8 bit parallel input down to 4 bits in order to save I O pins U6 is the microcontroller itself A 9 Building and testing the Mark One This design evolved over a period of a few months The various subsystems were tried out on breadboards in the lab and once they seemed to be functioning happily in isolation it was time to build a prototype and try out th
96. and the inverter The whole plant was equipped with data acquisition DAQ A dispatchable dump load was also developed 8 1 4 Mark Two software This plant served as a development system for the next generation of agent programs for economic inverter and battery management For the latter a rule based system was developed based on well known battery management rules used commercially An economic lead acid battery model was also developed which unfortunately could not be validated by experiments due to pressures of time and budget Inverter management was handled by an economic model based on stored efficiency curves of the inverter During the course of software development a number of unforeseen 126 issues arose which required a partial rewrite of the dispatching engine and communications layer Eventually the system was brought to an apparently satisfactory operating condition A datalogging engine was added and the system was considered complete and ready for evaluation Experiment design But how would this evaluation be done Evaluation conjures up an image of testing system A against system B to see which is better But in this case the system is a new invention There is no other system for small scale economic dispatching to test it against so any comparisons must necessarily be indirect perhaps with traditional automatic generation controllers What is more a comparative test like this is not necessarily meaningf
97. apacity seems to vary with the discharge rate with the battery seeming to run out sooner than expected at higher currents The actual capacity does not really change though because the unreacted material is still present so if the battery is allowed to rest for some time it may recover The precise relation between capacity and discharge rate is a complex function of the battery construction and also of time a battery which has been discharged at a high rate will recover to a higher apparent state of charge a while after the discharge is finished as the sulphate tends to redistribute itself more evenly The process has been modelled in various ways to high levels of detail 1 but a reasonable approximation can still be had by the 100 year old equation known as Peukert s Law which relates discharge current time and capacity I t C Eq 7 1 Or in terms of capacity substituting Q It QUESO Eq 7 2 In Eqs 7 1 7 2 I is current is estimated capacity t is time C and are constants found experimentally for a particular kind of battery C is related to the battery capacity typically is between 1 20 and 1 45 So by measuring the discharge current and using Peukert s Law an estimate of the available capacity can be found But the problem is not solved yet because the answer is in amp hours of charge not the available watt hours of energy To calculate this the battery voltage must also be considered 94
98. are not quite as resistant to damage by overloads 3 Given these advantages and disadvantages the actual device chosen will depend on what sort of inverter circuit is chosen this determines the voltages and currents imposed on the devices and on what control algorithm is chosen this determines the speed at which switching must be performed A 5 2 Circuit topologies The most simple and well known kind of inverter is as shown in Fig A 1 It consists of four switches which connect the DC supply symbolised by a battery across the output terminals first in one sense then in the other In this way the voltage is periodically reversed Voltage conversion is often required where an inverter is used This is the case in the present application where 24V DC must be converted to 230V AC The circuit of Fig 1 has a fixed output that is determined by the voltage of the DC source There are two common ways of circumventing this the simplest being to apply the AC output of the inverter to a transformer and so step it up or down to the desired 158 voltage This circuit is sho wn in Fig 2 more complex method is to change the voltage of the DC source instead by means of a DC DC converter This is made of an inverter sometimes called a chopper in this application followed by a transformer followed in turn by a rectifier as shown in Fig A 3 Although this is more complicated it has certain advantages When the inverter output
99. argued in this work that the best way of achieving this is by distributing it like the generators and loads it serves An omniscient central executive which solves the whole network for a global optimum requires data from the whole network and is therefore susceptible to malfunction if any data connection in the network should be broken Therefore the robustness of such a system is seriously in question and it is to 42 be avoided If generators and demands need more information to make their decisions they can get it elsewhere With respect to these considerations the proposed solution is thought to be a reasonable compromise for initial purposes at least There are also issues of satisfaction and perceived fairness These are embodied in Rule 3 where they appear as a trade off between supplying electricity to buyers who will only pay low prices which is necessary to generate customer satisfaction and making profit for the dispatcher which must be done otherwise there would be no incentive to operate the system This trade off is implemented by setting a profit margin as proposed earlier by following the rules set out the profit might be expected to increase rapidly as the dispatching proceeds reach a peak and then fall off again The dispatching process would be terminated just before the profit falls below the profit margin The higher this is set the more the system emphasises making money as opposed to keeping customers happy
100. batteries The system supplies power for the door entry systems and distribution amplifiers in three tenement blocks F 2 Controller requirements The controller had to function as an ordinary charge controller i e prevent the batteries from under or overcharging It also had to control the heat recovery fans so that they did not run when there was insufficient heat Battery charge control was to be achieved by turning on a dump load when the voltage became too high and turning on a backup charger powered by mains electricity when it became too low Fan control was according to the PV array current the fans were turned on when the current rose above a threshold F 3 Design considerations For various reasons which will not be covered here a decision was made to use a traditional analogue type circuit instead of a microcontroller A voltage or current controlled switch was designed and four units of it were used for battery charger dump load and two fans respectively 231 F 4 Theory of operation The basic function performed by the controller is to change the state of a switch according to the value of a voltage or current input To avoid unnecessary switching and to implement certain functions hysteresis is used in other words the value of input that turns the controller ON if it was previously OFF is made different to the value that turns it OFF if it was ON So for instance the controller might be set to turn a batter
101. bility for power measurement of the inverter can be delegated to the inverter itself If its power control is accurate it may be sufficient to assume that the power output is equal to the demanded power output The approach used was to calibrate the inverter using a laboratory wattmeter and store the calibration table in the inverter control software This will be discussed in more detail in appendix A 6 4 Setup and commissioning At this time most of the parts required for construction of the experimental system were ready The inverter had undergone several test runs and seemed to be stable and reliable The dispatching programs had been tested with sample data and a PC with eight channel data acquisition card was ready Batteries and PV modules had been procured The time was ripe to assemble everything and risk some preliminary test runs 6 4 1 Equipment list Saturn 80 monocrystalline PV module BP Solar 2 12V nominal 80W peak MSX 50 polycrystalline PV module Solarex 2 12V nominal 50W peak Inverter 24V in 230V out 550W max In house See appendix 1 Dump load 24V 200W max with shunt In house See appendix 2 regulator and control input Terminal board with fuses battery In house 1 isolator current shunts and voltage 1 8 PC Compatible computer 333MHz Various 128 RAM PCI 6023E data acquisition card 8 National Instruments analogue input channels 8 digital output
102. by a number of nodes interconnected by components each having resistance and reactance At run time the database of nodes and components is converted to an admittance matrix for solution As this description suggests the solution is in the form of phasors and there is no transient capability The possible kinds of connecting components are limited to sources of power sinks of power resistances and reactances PFS is integrated with the rest of ESP r in that sources and sinks of power can be tied to the building and plant simulator This means for example that a PV module modelled in PFS will change its electrical output according to the light intensity on it as calculated by the building simulator The opposite is also possible sinks of electrical power can be made to appear to the building simulator as sources of heat at the appropriate locations and times Furthermore the tie in with the plant simulator means that as electrical plant items are turned on and off by thermostat etc PFS will see the load changing in the appropriate manner While this integration is certainly very impressive it does not alter the fact that there is no way of modelling an inverter This is because an inverter is a non linear connecting component which cannot be represented as an impedance Rather it looks more like a power sink paired with a power source E 3 Active connections In order to deal with this PFS was modified slightly A new type of conn
103. c instruments connected to a computer Complex experiments can be carried out with far greater speed and convenience than if the instruments were operated by hand Within the University of Strathclyde applications for LabVIEW have included Monitoring and control of cogeneration plant Realtime control of a robot arm Sound and vibration analysis Weather datalogging Recording and analysis of electromyograms 240 G 3 Basics of LabVIEW programming The idea behind graphical programming 1 that programs are not written but drawn by placing graphical icons and interconnecting them with lines all using the mouse As an example we can take the famous Hello World program The Hello World program Include lt stdio h gt Int main void printf Hello world n return 0 Fig G 1 The Hello World program Hello World is given above as it would appear in a conventional text based programming language It can equally well be written or rather drawn in LabVIEW where it looks like this Ib helloworld vi Diagram C3 1 289 P Hello world Fig G 2 The Hello World program in LabVIEW 241 Lie Edt Opere Tools Browse window Heb ofa Sm Text display P Hello world Fig G 3 Output from the Hello World program Creating a LV program is really more like building a piece of electronic apparatus than writing a conventional program a delib
104. cept that an expanded scale is not required so the negative end of the meter is grounded F 5 3 System wiring Refer to Fig F 3 233 The complete system consists of four controller boards two configured as voltage controllers and two as current controllers one power supply meter board three meters two six way meter switches and four relays The third meter always reads Array 2 current It is not connected to the meter board but directly to the Array 2 current shunt input with a series resistor for correct scaling The relays are provided with flywheel diodes to clamp back EMF and contact suppressors to reduce spikes when switching inductive loads F 6 Modifications Meter scale resistors on the metering board were replaced with 10 turn trimmers to aid calibration of meters Another 10 turn trimmer was added to the voltage reference for adjustment of the voltmeter offset 234 19 JUMPER see note 25V see note see note Values of RI R2 R3 R4 set gain of ing RI R2 R3 R4 see note Voltage controller with range 10 15 m R3 R4 13k Current controller depends on shunt 7 14LM324 14 LM324 14 74HC132 TA74HCI32 1 4 LM324 1474HCI32 D3 INA148 1N4148 Fig F 1 Controller circuit 235 Q1 2N3906 Us NMEI2I2S ab 2N3904 Us 236 Power supply to controller circuits Voltage sense connections Voltage controller Current controller Current controller control
105. ch is a well known concept in digital logic systems see 8 It determines its current state based on the previous state and its inputs The state diagram is shown in figure 7 2 The only 106 volts 13 8 demand 0 amp volis 13 8 a volts 15 recycling state 1 d initialise 0 equalisin 0 9 9 elapsed 2 Idle state Neither buy nor sell power Buy state Buy any available power considering set price Sell state Advertise power for sale Hecycling Buy any available power with no selling allowed Equalising As recycling until battery voltage has been high for a few hours Initialise When program first run Fig 7 2 State machine for battery management 107 new feature is the initial state which was not specified in the above rules but which every real state machine must have This is the state entered as soon as the state machine is started by running the program It leads to the Recycle state so that the batteries will fully charge and the coulomb meter be reset to the proper state of charge 7 1 14 REDMan connection Connecting the system to the REDMan dispatcher proved to be something of an issue The main problem was this A battery being charged is not like a normal load such as a light bulb The light bulb always consumes say 60 watts On the other hand the battery will take whatever power is available Unfortunately the REDMan dispatcher had no way of un
106. chapter presents the findings of the initial literature review by way of background It looks at how energy is used citing statistics on global energy production and consumption with a view to finding out where energy is wasted and whether this waste is avoidable Out of various possibilities cogeneration of heat and electricity is selected as a promising way of reducing energy waste and investigated further Problems which prevent the wider use of cogeneration are identified and explored in greater detail 2 1 The use and waste of energy As a rough estimate the Earth s human population currently use 11 6 terawatts teraz 10 of all kinds of power in the course of their daily business 1 Of this approximately 3 comes from renewable sources and 7 from nuclear power but the remaining 90 is from fossil fuels Estimates of the size of fossil reserves suggest that natural gas might run out in approximately 20 years oil in 50 70 years and coal in perhaps 200 years From then on nuclear and renewable energy will be the only options As far as nuclear energy is concerned the risks associated with radioactive waste and nuclear plant accidents are well known The cost of providing safety measures against these risks combined with adverse public opinion has seriously hindered the nuclear industry For example 3 states that no new reactors have been ordered in the United States since 1979 So fossil fuels are running out and nucl
107. circuitry of their inverters or the computer firmware that controlled them even for academic purposes This is obviously wise practice in a commercial scenario but it would mean that modifying their product would be a matter of reverse engineering it in other words poring over the circuit boards with a magnifying glass and trying to reconstruct the firmware from the raw machine code extracted from ROM chips This would not only be a very difficult job but probably an offence under intellectual property law On the other hand while building an inverter from scratch might seem more difficult the whole reverse engineering issue would be avoided and more importantly the 155 plans could be placed in the public domain where they might be of use to other researchers A 3 Design objectives The most important objective and the whole purpose of the exercise was that the inverter must be controllable in real time by a computer It must accept commands telling it how much power to transfer from DC to AC at a given instant It would also have to be reliable It is all too easy to create a unit that performs OK on the test bench but fails in the field The unit would have to operate successfully for the duration of the experiment at least one month Efficiency was also a concern If the experiment was to give realistic results the efficiency would have to be representative of the efficiencies of contemporary commercial invert
108. control LFC which adjusts the mechanical power driving the generators to hold the supply frequency constant Alongside this is economic dispatching ED which aims to share out the total demand amongst the generators in such a way that they operate at the best overall efficiency These jobs are collectively known as dispatching and this will be the main challenge that this hypothetical EG network will face reconciling the traditional methods of dispatching developed on retworks with a few large generating stations with the new requirements of EG There seems to be no real idea of how to accomplish this As a source of inspiration though it may be instructive to examine the way in which dispatching is currently done 2 6 1 Alook at dispatching Dispatching was originally performed by a combination of simple automatic controls and people in a national control room in contact with generating stations by telephone Human dispatchers have since been superseded by computerised automatic generation control but the objectives are still the same and can be seen as consisting of three fundamental parts predicting demand coping with unexpected demand and trying to reach an economic optimum 2 6 1 Predicting demand The demand for electricity in each half hour period of the day is forecast one day in advance This is possible because there are daily and seasonal patterns in the demand related to working hours mealtimes television p
109. control centre calculating optimal setpoints for all the generators and then sending this information back This is feasible in current networks with 10 100 generating units but in a future scenario with perhaps millions of small embedded generators it may well be impractical For example cons ider the scenario where one in five UK households has replaced their domestic heating boiler with a small grid connected CHP unit Instead of 100 generators the control centre has to deal with 5 000 000 Data flow to and from the control centre will increase by four orders of magnitude but the worst is yet to come The amount of computing power required to run the control algorithms is typically proportional to the square of the problem size if not some higher power Therefore the computing power required 14 will increase by at least eight orders of magnitude put this in perspective by Moore s Law which states that the power of computers doubles every 18 months it is around 40 years worth of development Of course it might well be possible to manage such large amounts of data successfully But objections can still be raised to the general concept of collecting all data together to a central executive The chief objection is that it represents a serious reliability issue If the control centre broke down the whole network would be rendered useless So it is a weak link in a system that should be strengthened by redundancy Also becau
110. cost in terms of computing power 6 5 Summary Hardware design issues related to grid interconnection and component choice were investigated An experimental test bed consisting of four PV modules two wind turbines a 24 volt lead acid battery bank a grid connected power inverter a dump load and a data acquisition control computer was assembled and commissioned successfully The inverter and dump load were designed and built especially for this 90 project See Appendices A B as also were two different kinds of demand sensors Appendices C D The next step is to develop and test the agent programs that interface the experimental hardware to the dispatching system 6 6 References 1 Electricity Supply Regulations 1998 HMSO 1998 Part VI Section 26 and Schedule 3 Available online at http www hmso gov uk si si1988 Uksi 19881057 en l htmfitcon 2 ADE7755 Energy Metering IC with Pulse Output datasheet Analog Devices Inc 1999 Available online 91 Chapter 7 Agent software This section covers the development of agent programs The term agent is used here to denote programs which interface between actual energy storage conversion hardware and the dispatching engine The basic philosophy and assumptions as to the roles of agent and dispatcher were developed in chapters 2 and 3 To recap In the case of a generator or other source of power its agent informs the dispatcher of the maximum possible pow
111. could be labelled or weighted according to its type e g renewable fossil cogen etc This might override the economic dispatching process enabling environmentally conscious consumers to select their desired type of energy instead of being supplied with the cheapest by default This would be a further extension of the combined heat and power dispatcher proposed in 10 5 7 However the complexity of such a system may prove to be prohibitive 10 5 10 Strategic barriers to implementation So far the problems discussed have been purely technical ones such as making the system perform with large numbers of generators resolving specific bugs and issues and so on However there are also some broader strategic issues which have been touched on in passing throughout the text but which deserve to be summarised here The first is that of dispatchability Most EG plants as they are designed and built today are not able to receive dispatching commands from a system like REDMan Therefore existing EG plants would need to be replaced or modified and the design of new EG plants would need to be changed The second barrier is a lack of co ordination between EG and the existing control and protection systems such as frequency control and antr islanding The protection systems currently used in power distribution networks may well not be compatible with REDMan like systems Also the protection systems currently used in EG plants may lave negat
112. cs and so was eliminated by forcing all currents less than zero to zero However if this had been a genuine leakage current it would explain the error It would also account for part of the missing energy from the battery which delivered three times more energy than the dispatcher credited it with In order to answer this question the DAQ system accuracy should be checked The dump load dumped more energy than the dispatcher ordered it to This could be explained by the voltage regulator function built into it which clamps battery voltage at 30 0V If the battery was temporarily overcharged due to slow response of the battery management program the voltage regulator would dissipate power to keep the voltage under control This hypothesis could be checked by examining the datafile in more detail see section 9 4 Transient behaviour 9 3 DAQ Hardware check DAQ current and voltage readings were checked against a 3999 count digital multimeter This 1s no better than the 12 bit 4096 count DAQ board however the meter has a wide choice of ranges whereas the DAQ is limited to one fixed range This effectively makes it much more accurate than the DAQ when dealing with currents that are a small fraction of the DAQ s full scale FS The DAQ and meter readings of battery voltage agreed to within 0 01 volt Some errors were noted on the current channels a constant offset of 0 03 A on the 10 A current channels except the dumpload which ha
113. ctioning according to a specification Of course creating the specification was part of this project too and at the beginning it was only the vaguest of concepts something that would make renewable energy systems more cost effective or more useful through intelligent control It was through experimentation trying out various ideas writing different kinds of control software and so on that this vague concept evolved into a real machine that actually did something This development work has already been covered in previous chapters and will only be briefly summarised here Rather this chapter is concerned with the second kind of experiments those which test the performance of the whole system to see whether it is good or useless and whether the project has been a success or a failure Obviously in order to do this a standard is required to judge it against because goodness and success are not absolutes So the question is what experiments should be done to prove that the system is good This is a serious question which must be tackled before any experimental design is attempted Summary of development work so far Software The first task undertaken was to write a computer program that performed dispatching according to the simple rules specified in Section 4 3 1 This was successful Next was to develop a communications infrastructure which would connect the dispatcher to the power sources and demands for which it was r
114. cuit whose output is HIGH if the instantaneous value of the modulating wave is greater than that of the carrier wave and LOW otherwise The result is a train of pulses 161 repeating at the carrier frequency with the width of each pulse proportional to the value of the modulating wave A spectrum analysis of this waveform would show that it contained a component at the modulating frequency a component at the carrier frequency and the harmonics of the carrier frequency This pulse train is used to operate the inverter s switches so that a high power replica of it emerges from the inverter s AC output terminals A low pass filter is then used on this to remove the carrier frequency and its harmonics while letting through what turns out to be a very good reconstruction of the modulating wave See 4 for more information The task of the low pass filter is eased by making the difference between carrier and modulating frequencies very large By using MOSFET switches which perform very well at high frequencies it is easily possible to have a carrier frequency of say 50 kHz and so the carrier can be greatly attenuated while the desired 50 Hz component is unaffected using only a simple second order filter In this way it is theoretically easy to meet the distortion spec The PWM generator can be simplified even more by using hysteresis aka bang bang control which is a technique borrowed from commercial inverters that drive i
115. curate and not very robust to errors in various components The datalogging should be redesigned with more redundant measurements and greater measuring accuracy The inverter should be improved perhaps with higher efficiency but certainly with accurate onboard measuring control of input and output currents and powers Safety regulatory issues associated with inverters should be reviewed since these are subject to ongoing change normally for the better Batteries should be investigated in more detail in order to confirm deny assumptions made in the battery management software Also policy on energy storage systems should be revised Systems without any storage at all might be investigated or other storage systems with smaller penalties for partially charged operation For example Some battery chemistries e g nicke cadmium nickeliron nickel metal hydride lithium ion suffer no ill effects from being left discharged Exotic technologies like flywheels and ultra capacitors are also eligible 10 5 2 Software issues Bugs in the inverter management were found These were not fatal but contributed to inaccuracy These should be removed as far as possible however their existence is partly tied in with incremental cost and multr domain dispatching 10 5 3 Marginal cost A dispatching algorithm without marginal cost optimisation proved to be adequate for controlling the experimental system However as a result of this research it was
116. d 0 01A and 0 09 A on the 30 A channel This corresponds to a voltage offset of 0 03 0 075 10 225 V in 136 either case which is within the card tolerance of 345 V 1 when operating at 500mV FS sensitivity The card has a higher gain setting with lower offset but it is not usable because the maximum input at this setting 15 only 50mV and signals of 75 150mV are expected from the current shunts The 0 03A error at 24V represents a power error of 0 72W on each 10 measurement channel and 2 16W on the 30A battery channel The renewable energy on 4 channels will be in error by 4 0 72 2 88W Thus 2 16 2 88 5 04W in total are unaccounted for The error on the dumpload channel should be reckoned as having the opposite sign because the dumpload controller regulates the DAQ measured current to zero thus the actual current will be the same as the DAQ offset but with the opposite sign Therefore the dumpload measured energy will be less than the actual amount Over the 91 67 hour duration of the experiment the errors in watt hours work out as Measured energy Error Wh Corrected energy Wh 264 follows Table 9 3 The Error column figures are estimated from the offset figures calculated above and used to estimate what the energy flows would really have been To clarify the sign convention The R E actually generated more energy than the DAQ system measured The battery output more the dumpload dis
117. d p 508 Reactive power generation and control by thyristor circuits Gyugyi L IEEE Trans 1979 IA 15 5 521 532 Power Electronics Lander C W McGraw Hill London 1993 pp 306 311 Embedded Generation Jenkins et al IEE London 2000 p 106 ibid p 15 Development and Testing of an Approach to Antr Islanding in Utility Interconnected Photovoltaic Systems Stevens J et al Sandia National Laboratories 2000 Power Systems Stability Handbook Pansini A J Fairmont Press 1992 Electric Power Systems Nasar S A et al CRC Press 1998 p 88 Dynamics and control of large electric power systems Ilic M Zaborsky J Wiley New York 2000 Murphy s Law can be stated as If it can go wrong it will Basically an alternative statement of the Second Law of Thermodynamics Author unknown Mult Module Parallel Small Battery Energy Storage System Chiang S J et al IEEE Trans Energy Conversion Vol 11 no 1 pp 146 154 Line conditioning system with simple control strategy and fast dynamic response Moran L et al IEE Proc Generation Transmission amp Distribution Vol 142 No 2 pp 128 134 Reduction of voltage violations from embedded generators connected to the distribution network by intelligent reactive power control Wallace et al IEE Conf Publ No 488 April 2002 35 22 A variable speed wind
118. d but O n This is a serious discrepancy between theory and practice but can perhaps be explained by the way in which data is cached in the computer s memory Assuming a small cache of higher speed memory which is not quite big enough to hold an entire array the probability of accessing each array element being equal i e random accesses or sequential accesses over entire array the mean number of cache misses requiring a slower access to main memory will be proportional to size of array size of cache and hence the mean speed of performing operations on such an array will depend on its size Thus it is possible for an algorithm containing a number of operations of order to show an execution time of order n This argument no longer holds once the array is large compared to the cache This hypothesis can be tested by re running the speed test but modifying the program under test to include a fixed delay large compared to memory access time with each array access The effect of this is to make the access speed independent of the array size and the resulting execution times are O n In any case for small problems such as will be encountered in this experimental work the speed is quite sufficient For a 10x10 problem the program executes in 0 62 ms and even for a 100x100 problem it is still only 110 ms which is small compared to a 1 second timestep 5 5 Validation This is a less clear cut subject What ar
119. d and discussed here The experimental system was dispatched in a satisfactory manner and much practical experience was gained in the issues relating to dispatching of EG Several possible avenues for further research were identified vi CONTENTS 1 rinii I CHAPTER 2 BACKGROUND eee ennt toten essa sees 3 2 1 THEUSE AND WASTE OF ENERGY eene nnne rennen enirn enne then tns 3 2 2 GENERATION AND TRANSMISSION OF ELECTRICITY c ccccccsssssscssesssscsescescssescsecsesecsececsecsesecseceesecaeene 6 2 3 THE NEW IDEA EMBEDDED GENERATION eee ennt ennt tnnt tatnen testet attenta tan 8 2 4 THE TECHNOLOGY eT ER GER TOR 9 2 55 CONNECTING EG TO THE GRID ertet acne or atn e eei reete eU EE 11 2 6 DISPATCHING th ai eed on d THER 12 2 6 LI Predicting eR E 12 2 6 2 Coping With the unexpected a e e RE E ERR ER 13 2 60 39 ECODnOTRICS t ERN E E 13 2 6 4 ReneWables i ee oe eed eg e aide idee i n He PU 13 2 7 THE FUTURE OF DISPATCHINQG erret 14 2 7 1 Alternatives to classic eese 15 2 85 CONGEUSIONSz iater ERR tbi iO ER FRU rb e a HER HESSE 16 REEERENGES 16 CHAPTER 3 POWER SYSTEMS CONTROL
120. d in all other agents The load routine was used here it takes a desired power and price and returns a true false value authorising the load to switch on or not In this application the Opportunistic flag was set This modification to the basic REDMan dispatcher was discussed previously it means that if the load s desired power is not available it will be given whatever power is available rather than nothing at all When the load communication routine is operating in this mode the true false output is replaced by a number stating how much power the load can have Connecting this to a real dump load is a relatively simple matter The dump load hardware can be thought of as a variable resistance controlled by an 8 bit digital output from the computer When this is all 1s the dump load draws maximum current This is not very well defined and depends on battery voltage and temperature It was measured at about 10 amps When the output is zero the dump load draws zero current At some ill defined value between zero and 255 it starts to conduct Because of this vague characteristic the control program must use feedback to make the current accurately equal to a desired value The computer measures the dump load current anyway so the required data are present The feedback was implemented as an integral controller algorithm see figure 7 7 The control algorithm described above takes a current as input whereas the output from the communicatio
121. derstanding this it could only work in terms of definite powers This was a serious problem which required rewriting the dispatching engine The result was the Opportunistic flag This is a true false input which appears in the communications routine used by loads When asserted the dispatcher treats the power value sent by that load as if it were a maximum and if that amount is not available it sends whatever power is available In calculating the available power the price paid by the load is of course taken into account It operates on a per load basis so that loads using the Opportunistic flag function can be dispatched alongside loads without it Once this modification was made the rest was simple enough The powers sent to REDMan communications routines were chosen according to the state of the state machine All powers were either zero or a high value larger than any foreseeable demand in the system such as 1000 watts Fig 7 3 shows a functional diagram of the program A listing of the LabVIEW program is in Appendix H sections H 2 1 H 2 2 7 1 15 Testing BatMan Initial testing was performed in the actual experimental system under variable conditions of supply and demand Once debugged the program was observed to behave in a very similar way to an ordinary charge controller This was exactly as would be expected since it was modelled on one in the first place The testing and results will be covered in chapters 8 and 9
122. ds 5 1 dispatcher In chapter 3 a simple dispatching algorithm was proposed consisting essentially of just three rules buy cheapest electricity first sell to highest bidder stop when customers satisfied demand exhausted or operation no longer economical whichever occurs first But these rules are of rather a high level of abstraction and in order to incorporate them in a computer program they must be broken down into simpler elementary operations Providing cheapest commodity highest bidder functionality in support of the first two rules is easy enough The first step is to gather together the power and cost information for each source into a list of power cost pairs and do likewise for each demand A simple sorting algorithm to place these in order of cost cheapest first for sources and highest bidder first for demands is all that is required The source list and demand list are then fed to the dispatching algorithm proper It takes the first demand in the list and makes it up using power from the source list again starting with the first item In the meantime it also calculates the total cost of the energy taken from the source list and compares it to the price the demand 1 prepared to pay If the cost is greater than the price which will be paid the transaction is deemed uneconomical As soon as the first transaction of this kind is detected the dispatching process is finished because the order of sorting of the
123. dware software agent 120 LE ud S DataSocket LE ud e server EM Generation agent Battery Ethernet agent Inverter agent Computer e PIE engine Ethernet Load manager Fig 7 11 Software configuration of experimental system 121 7 6 Summary A set of computer programs were developed to interface the experimental system s hardware to the dispatching engine The complete suite of software is shown diagramatically in Fig 7 11 Selected programs are listed in Appendix H and the source for all programs can be downloaded from http www esru strath ac uk In the case of the renewable energy generators and the dump load there were no real problems On the other hand dealing with the inverter and battery was more complex The agents for these had to incorporate economic models for the costs incurred by running the equipment and also decision making logic which amounted to a simple form of rule based intelligence In the battery management agent the model and rules were developed from well known battery management techniques used commercially In the inverter management agent they were developed from first principles In every case they were refined by testing on the experimental system until performance was satisfactory These experiments will be described in more detail in the following chapter 7 7 References 1 Modelling battery charge regulation for a stand alone photovoltaic
124. e 1 000 MW generator as opposed to 1 000 one megawatt units power from the smaller generators should have more availability This is especially so for demands less than 1 000 MW which could be met even if several small generators were broken down In other words there is improved availability due to diversity and also possibly due to improved reliability of individual units 3 7 5 EG might improve stability In the dream scenario where every EG unit did its share of all the control duties such as LFC and AVR and all were economically dispatched power quality might be better than at present The number of degrees of freedom for voltage regulation This has been found in studies of larger scale generators e g 500MW vs 2000MW but there are no data for the case of very small generators 3l would increase so that voltage could be controlled more tightly everywhere Frequency control and transient stability might improve too simply because smaller units respond proportionately faster In fact power electronic inverters can break this speed size tradeoff altogether no matter what the size they are controllable with sub cycle precision 19 20 show good examples of such systems And lastly deliberate islanding could make for more redundancy and so reduce the impact of faults compared to today 3 7 6 Or perhaps not In 3 7 5 above evidence was presented to suggest that distributed embedded generation could improve p
125. e cost of electricity is the nominal 20 hour capacity and Vnom is the nominal voltage Now that the model is complete it is time to consider how the information that it needs might be gathered The data needed are Depth of discharge Maximum depth of discharge Estimated time to recharge Estimated price of recharging Depth of discharge can be estimated by one of the capacity measuring techniques investigated earlier The chosen method is the coulomb meter because it does not depend on properties of the battery which would have to be measured experimentally Maximum depth of discharge is easily derived from this using a peak detection routine The estimated time to recharge can be found by averaging the incoming energy over the previous week It is also time to consider what can be done with the output of the model This is a cost representing how much the battery wear and tear will cost if discharge was stopped at the present time and the battery recharged completely at a rate similar to previous charge rates logical step would be to calculate the profit returned by the battery How is profit defined The income of electricity sold since the beginning of the cycle less the expenditure of electricity bought These figures are easily available from the REDMan engine Subtracting the modelled wear cost from the profit gives the expected profit if sales were stopped immediately and the battery recycled A spreadshe
126. e et ye re e de i e e t e dy ts 131 CHAPTER 9 RESULTS OF EXPERIMENTS eeeeeeeeeeeeeeeeete 133 9 1 TEST RUN FOR ENERGY BALANCE TRANSIENT esee eter enne 133 9 2 ENERGY BALANCES 135 9 3 DAQHARDWARE CHECK cete ere ehe 136 9 47 INVERTER RE TEST ined AEN RARE EE PER AERE EE ATI be 138 95 TRANSIENT BEHAVIOUR 140 9 5 1 Typical dispatching 140 9 5 2 Actual dispatched power flows 5 142 9 5 3 In search of missing 142 9 6 FWO DOMAININTERACTIONG 145 SUMMARY 145 9 8 REFERENCES 146 CHAPTER LO CONCLUSIONS 147 10 1 DOES IT DISPATCH SUCCESSFULLY ccssssssssscsecsecssescessesseseesecsssssssssssessecesceseeseacenecsecsecsaesaeaaees 147 10 2 DOES IT WORK ECONOMICALLY c ccccsscsssssessecsecsecssesseseeseeeesecsecessesesseseecsceecssessesecsecaecsseseeneees 147 10 3 WILL IT SCALE TO LARGE NUMBERS cccssscsscsscssessessecsecsecsessesseseesesesesceeseseeseeaecsecsecsaesaeneees 148 10 4 OVERALL CONCLUSIONS nnee areara RER E E EA E E EE iA 148 10 5 FURTHER WORK 149 10S l HardWare 149 10 5 2 SOffW re ISSUES s cce trei e EA ede RR 149 105 3 Marginal Costi HR RR SUES Rg a EE eee ES 149 10 5 4 M lti domain
127. e the criteria by which the efficiency of the dispatching should be measured There has been other work such as 2 3 which attempts to quantify dispatching efficiency by metrics such as the amount of control movement and the network frequency stability due to a given algorithm However as argued in chapter 4 the goal was not to optimise either of these the first is assumed to be irrelevant due to the response speed of small generators and the second delegated to lower level automatic controls The proposed dispatcher is 62 concerned only with economics and under the simplifying assumptions made earlier the dispatcher s optimising problem is fairly trivial It is only in a situation where at least one source has nomzero incremental cost that there is anything worthy of serious mathematical optimisation Doubtless a more advanced algorithm could be specified but a case has already been put forward in favour of the simplest approach possible There are many scenarios renewable energy storage devices domestic grid electricity where the price of the product can correctly be assumed independent of the amount bought because it is too troublesome to implement price breaks In these situations the dispatching is indeed trivial from a mathematical viewpoint but the technical and organisational issues involved of actually measuring and directing the power flow according to the commands of the dispatcher are far more serious This wo
128. e was switched off it would notify that the ID was no longer in use However this protocol is relatively complicated and in the first instance where there might only be 10 agents in total it would be no great trouble to allocate the numbers statically by hand The problem of opening and closing connections can be avoided thanks to a peculiarity of DataSocket discussed earlier an item once written to the server is not altered by being read By arranging for agents to write a power supply or demand amount of zero when they quit the receiving end of the DataSocket connection will keep returning zero every subsequent time it is read As far as the dispatching algorithm is concerned a source with zero maximum power or 68 load with zero demand is ignored completely This might seem like a quick fix computing power is being wasted to read connections every timestep when they carry no useful data A proper allocation system which obliged the agent to request a connection would be more economical Again though the speed tests showed that there will be no shortage of computing power Relying on the agent to close the connection in this way may also reduce the robustness of the system for instance if the agent crashes or is terminated before it can write a final zero the last value that it managed to write will persist So it is evident that a proper handshaking allocation procedure would be very important in a future real sy
129. e whole system The emphasis at this stage was not on perfect performance or error free operation but simply to provide a proof of concept A printed circuit board PCB was made and stuffed with components and the various stages of the circuit were tested in isolation before connecting everything together Inevitably a number of design flaws were discovered and changes had to be made to the circuit 177 02 301 A01 A01 ZHI 058 6 9 ds 1 e En sn 851 lt 84 C id lt lt ed lt ed V asi troller and support components ircuit of microcon 12 Fig A 178 Fig A 13 Mark One inverter under test Volts Div Coarse Fig A 14 Line current of Mark One operating at 100 output Y Scale 200mA div 179 A 10 Lessons learnt from Mark One A 10 1 Odd spikes The most puzzling anomaly in the Mark One s operation was a strange disturbance in the current waveform When the switching devices change state the rate of change of current is supposed to reverse This it did but accompanied by a very large transient a spike which disturbed the bang bang control system quite severely To allow proper operation the spikes were reduced by low pass filtering the current signal but the true cause was discovered quite by accident If it is to work correctly the i
130. ear energy is unlikely to step up to take their place Therefore future energy scenarios are likely to involve reducing demand and increasing the amount of energy drawn from renewable sources Reduction of demand will be considered first An excellent way of doing this is reducing waste of energy so it is logical to ask the question Where do the greatest wastes of energy take place Detailed studies e g 1 2 have been made of energy use patterns and some representative figures are presented here These figures are average power flows over a one year period 3 Of the 11 6TW consumed as primary fuels 5 TW 43 1s used in electricity generation which is approximately 35 efficient so 1 75 TW of electricity is the result 1 4 2 TW 36 is used for heating of buildings domestic hot water and industrial processes 2 1 9TW 16 goes to transportation of people and goods 2 The vast majority of this is of fossil origin Motor vehicle engines convert about 20 of the fuel energy into mechanical work Of course the question of what proportion of this energy is actually wasted is a difficult one In transportation for example not all journeys are necessary In heating the insulation of buildings can always be made better And there are many uses of electricity which might not be considered indispensable Finding answers to these questions would require a detour into politics economics and the social sciences Bu
131. ecting component was introduced the active connection This was made from a combination of existing PFS components One node was a power sink the other a power source and a very high resistance equal to the largest number representable by the computer was also placed between the two nodes in order to keep the solver happy The power sink and source were connected to a new piece of code which modelled the inverter itself The configuration of power sink and source would depend on the kind of inverter being modelled 226 The initial case tackled was a grid intertied inverter without battery capability In operation this device simply takes the maximum possible power using MPP tracking from the PV modules and dumps it into the grid Therefore the start node should be represented by a power sink that eats all the power available at a node The end node should be a power source that emits whatever power went into the start node less the losses in the inverter The main trick in this respect was specifying all the power available at a node PFS has a few different node types It is possible to have the node voltage fixed or floating In the case where it is fixed the solver imports or exports extra power to the node as needed to preserve the energy balance The fixed voltage node is normally used to model the power source in an electrical network because it looks like an ideal voltage source For this same reason it also functions as a
132. ed by PORTB unless not in run mode movf H byte W movwf PORTB goto noclr if not in run mode send zero instead clrt PORTB increment wave step index for next time testing for overrun noclr movlw ISAT value to saturate index to subwf index W btfss STATUS Z if zero flag not set inet index F increment index all done clrwdt keep watchdog timer happy retfie vetfie also re enables interrupts e ck ck ck ck KKK KKK KKK Ck Ck ck ck ck ck ck ke x KKK KKK POWER ON INITIALISATION e ck ck ck ck ck kk kk kk kk Ck Cc ck ck ck ck ke Sk KKK KKK initialise working registers init bcf STATUS RPO bank 0 clrf power clrf index clrf errf set up initial values clrf movlw TMR movwf errc for timer speed movlw CTR movwf run countdown to start movwf fvi movlw TRIPPED movwf pstat protection algorithm status setup timer bsf STATUS RPO bank 1 bcf OPTION REG TOCS timer mode not counter bsf OPTION REG PSA prescaler to watchdog timer bcf OPTION REG PS2 set for 1 1 prescaling bcf OPTION REG PS1 gives nominal 18ms WDT bcf OPTION 50 clrwdt 203 bet STATUS RPO bsf INTCON TOIE enable timer overflow interrupt clrf TMRO preload timer setup PORTA movlw B 10000 output latches to known state movwf PORTA RA4 1 to hold power stage disabled bsf STATUS RPO bank 1 moviw B Q1111 all inputs TRISA 1 except RA4 movwf TRISA Setup PORTB
133. ed by the power demand from the loads or the power available from the generators storage system whichever is the smaller It should be possible to achieve this effect by appropriate logic in the inverter control program which will be considered later but this is not the whole story there is still the matter of actually making the inverter output the amount of power desired by the program This will mean having it dispatchable under control of a computer and unfortunately there are no inverters available on the market which can do this They all have built in maximum power point tracking Therefore it would be necessary to modify a commercial inverter or build one Building was chosen as probably the easier option A 550 watt grid intertied inverter was designed and built as part of this work Details of design construction and testing are in Appendix A 6 3 Considerations in setting up the system Now that decisions have been made on what items of plant to use it is time to consider how they should be connected together and how they should be interfaced with the computer running the dispatching functions Defining the basic hook up 1 easy enough The system will have as it were two buses a DC bus and an AC bus All DC plant PV DWTs dump load batteries connects to the DC bus and all AC plant loads grid connects to the AC bus The inverter connects to both buses and links them together 77 The hook up of measurement equ
134. ed to be rather a complicated system and moving data through it at a reasonable speed might therefore require a lot of computing power In order to find out more some tests were called for 5 6 3 Testing DSTP DataSocket A few small experiments were done with DSTP and the DataSocket suite to determine its suitability for the project Initially a rough estimate of speed was made by using a simple program that started a timer wrote an item to a datasocket immediately rad it back and then stopped the timer This gave the time elapsed between starting to place an item on the DataSocket server and finishing to read it back i e the latency This was quite variable it was observed to reach under lms but could also occasionally be up to 20ms The mean latency was about 0 5ms These figures were obtained when running the DS server on the same computer as the test program when communicating with a DS server on a remote machine via Ethernet they were somewhat higher if nore consistent a mean of 3ms with occasional peaks up to 6ms This is the time taken to transfer one double precision floating point number The REDMan application involves sending a series of values using one DataSocket write after the other It is reasonable to suppose that the result could be scaled up by multiplying the time measured above by the number of values which are to be sent 65 Another parameter of interest was the overhead i e the amount of extra data
135. ef PORTB now portb 0 we can test that DC input is ok this is dodgy we hooked a comparator to one of the port pins normally used as output ie time multiplexing bsf STATUS RPO bsf TRISB 1 briefly turn rbl to an input bcf STATUS RPO nop wait for things to settle nop nop btfss PORTB 1 if the DC is out of spec this 1 goto dctstok So if not 1 skip to ok movlw TRIPPED movwf pstat Status tripped moviw CTR reset fvi movwf fui dctstokbsf STATUS RPO bcf TRISB 1 bef STATUS RPO wave bcf return the hard way resetting the program and eventually overflowing the stack but this is not a problem since the stack is a circular buffer bsf INTCON GIE goto main timer interrupt service routine living dangerously we ENABLE interrupts when executing this code F INTCON TOIF clear the interrupt that got us here bsf INTCON GIE interrupts on first calculate the timer value this needs explained the timer value is made from the coarse error plus a simple dither arrangement the timer is bumped up by one if the current step index is less than the fine error if you think about it this controls the period in increments of 1 128 of a step we compute errf index if this is or 0 then the timer is incremented note the higher the error values the FASTER the timer will go movf errc W Save errc into timer movwf timer movf errf W bcf STATUS C subwf index W errf index btfsc STA
136. eing a 120 year old technology their behaviour is very well understood and characterised if not entirely simple And unlike the more esoteric technologies they can be bought cheaply over the counter However they have 92 certain odd behaviours which will need to be taken into account when designing the management system 7 1 2 How lead acid batteries work The operating principle of the lead acid battery is well known It consists of a positive plate made of lead dioxide and a negative plate made of metallic lead immersed in an electrolyte of sulphuric acid As the battery is discharged both plates are converted to lead sulphate and the acid concentration of the electrolyte decreases Upon recharging the plates return to their original composition This seems simple enough but in practice there are a number of peculiarities which make things more challenging The most exasperating feature from the energy metering point of view is that the capacity and efficiency of the battery is not constant but varies in a complex way To understand this behaviour a more detailed examination of battery physics is called for First of all what is meant by capacity From an energy systems point of view it would be defined as the amount of energy contained in the battery which is available for use This is the quantity that the control algorithm will need to know in order to make its decisions But how can it be derived from the battery
137. elemental areas N in the battery is large 2N 9 La EE dt N Eq 7 4 Then remembering that an area once sulphated stays that way the above expression can be integrated to find the fraction of the battery s plate area which will have become sulphated over the period of time in question Substituting 7 3 into 7 4 and integrating with respect to time 100 P9DOD Odt Eq 7 5 where C is a constant equal to the fraction of plate already sulphated at 0 Note Since this integral is the sum of a large number of small areas the overall DOD can now be other values than or 1 Now when the plates are completely sulphated the battery will be useless and will need replacement In this case equation 7 5 will be equal to unity and the financial penalty will be equal to the cost of the battery So multiplying equation 7 5 by the cost of the battery completes the model P and N may be lumped into one empirically derived constant TTF which may be pictured as the time taken for the battery to wreck itself completely if it was left totally discharged 5 opop yit TTF Eq 7 6 The shedding of active material must also be allowed for Some research has been conducted on this mostly by battery manufacturers desiring to improve the life of their products Published data 7 suggest that the amount shed is proportional to the total integrated discharge capacity This could be quite complicated to deal
138. ell known that existing schemes for planning dispatching and protection of central generators are not directly applicable to the new technology and dispatching scheduling of small embedded generators is not currently feasible In this work a novel dispatching management system which may meet this requirement is proposed Instead of using a single control centre it distributes dispatching functions throughout the network Some functions are performed by controllers and software agents built into the embedded generators themselves and others handled by dispatcher software associated with a group of generators and loads The dispatcher operates a small virtual market where energy can be traded between agents representing generators loads network functions AVR etc and other dispatchers This allows multiple dispatchers to be interconnected so potentially dealing with very large numbers of generators To test this concept some prototype agents a basic dispatcher and means of communication were created in the form of programs on a desktop computer The REDMan suite of software achieved successful trading of energy in a simulated environment This motivated a more advanced trial where REDMan was developed further and used for experimental dispatching of real generating equipment and loads Construction and assembly of the experimental apparatus interfacing of hardware to the computer environment experiments and results are presente
139. em with the mult domain concept as such This suggests that variable time stepping should be rationalised and the protocol for inter domain connection looked at A more detailed discussion of this issue is in 10 5 5 and 10 5 6 below 10 4 Overall conclusions The experimental program set out to achieve the following demonstrate dispatching of embedded generators using the software developed show that it was working in an economic manner and show that it was usable with large numbers of generators and loads The first objective was achieved the system was shown to dispatch successfully This in itself appears to be a new achievement in the EG field As for the second economics this was handled adequately considering that it was partly overridden to allow the inclusion of battery storage Some power was wasted compared to the optimal case but this was due to measurement errors rather than any kind of faulty dispatching Finally the most contentious issue scaling to large numbers remained open The system functioned satisfactorily with two dispatchers working together However transient response was slower than with one dispatcher working alone It was felt that this could have been remedied to a great extent by revision of the inverter 148 management and rationalising of the various modules settings for variable stepping 10 5 Further work 10 5 1 Hardware The experimental hardware proved to be rather inac
140. ent in the secondary will be nearly 50 A So it is possible to calculate the resistance of the secondary that will drop 2 5 of 12V 300mV when passing 50 A It is 0 006 Next knowing the core dimensions the length of one turn can be calculated 0 18m Therefore 23 turns will use 4 14m of wire The resistivity of copper is 1 72 x 10 m so if it is to have a resistance of 6 n a conductor of this length must have a cross section of 1 19 x 10 6 nf or 12 mm A single copper wire of this size would be very difficult to handle so a number of smaller wires in parallel is preferable 2mm diameter wire is easily available and using four strands of this gives a cross section of 12 6 mm However there appeared to be plenty of room on the core and so it should be possible to be conservative and use five So the existing secondaries were unwound and replaced with five 24 turn windings of enamelled copper wire 2mm diameter connected in parallel 183 A 11 6 Filter components Say that the switching frequency is not to exceed 30 kHz Now it was observed from the ATP simulation and experiment with the Mark One that the highest switching frequency occurred when the instantaneous line voltage was about half its peak value that is 12V Eq A 7 Now say the system is running at a hysteresis of 596 of the peak current 3 5A One half switching cycle can then be calculated as the time taken for the current to change by 3
141. er Provision was made for the integrator to be reset by an external signal The coulomb meter employed correction by Peukert s law according to the magnitude of the measured current so that it could give an estimated depth of discharge The wear cost model equations were also solved once per timestep using the depth of discharge from the coulomb meter program both normal and peak detected The peak detector was to be reset by the same signal that reset the integrator The output from the energy bought sold routine was subtracted from the wear cost to give the expected profit updated every timestep Executive decisions in BatMan were taken based on the expected profit and also on the battery voltage relative to float and equalising thresholds and an equalising timer The basic course of action according to the battery management techniques discussed earlier in this section was this Allow the battery to buy or sell energy unless The voltage goes above the float threshold in which case sell only The expected profit falls below zero in which case go into recycle mode In recycle mode buy energy only When the voltage rises above the equalising threshold start the equalising timer When the equalising timer finishes reset all integrating registers in the program assumed that the battery is now totally full stop buying and return to sell mode These rules were easiest implemented in software as a state machine whi
142. er it can generate and the per unit cost of this power The dispatcher replies with the actual power that it requires of the generator In the case of a load its agent tells the dispatcher how much power is required and what per unit price it is prepared to pay The dispatcher returns a yes no answer as to whether the load may run or not 7 1 Abattery management agent REDMan divides network entities into two classes energy sources and energy demands However it would seem that an energy store does not fit this classification because it could be either depending on whether energy is entering or leaving it This is not a problem because a single energy store can be represented as two separate objects in the REDMan world an energy source and an energy demand The problem is rather one of deciding just when the energy store should buy energy and when it should sell it and indeed when it should do neither It was argued earlier that this is not REDMan s job it simply acts as a trading centre where supplies and demands are matched The business of how much to supply or demand and when to do it is left to software agents acting on behalf of the loads and generators So the task would be to create an agent for a battery storage system 7 1 1 Why batteries There are many energy storage technologies from compressed air to flywheels and ultracapacitors But lead acid batteries recommended themselves straight away for this application B
143. er of PV modules used for other research projects in the department which could be borrowed Once fixed to the roof they could be properly classed as embedded renewable generation pair of experimental ducted wind turbine DWT modules were also available and there was some interest in how they would perform in a system of this kind so the intention was to use these also Unfortunately while the use of PV is commendable from an environmental viewpoint it causes some new problems of its own The first of these is that PV in its native form is not dispatchable The electrical output is determined by the insolation and also the temperature to a certain extent If the dispatching program demands more or less power than the PV can supply something must be done The second is that PV modules supply DC whereas the electrical grid and the vast majority of loads require AC The DWTs present exactly the same problems 6 2 1 Dispatching PV wind energy Dealing with the situation where the PV supplies less power than required is easy Recall that the dispatching algorithm receives a maximum power value for each generator If this is set to the maximum possible power given the insolation at that 75 timestep then the PV will never be asked to supply more power than it can Slightly more involved is the scenario where the dispatcher asks for less power than the PV is supplying Here some electronic intervention is required by diverting the surp
144. eral contracts In this the utility is unbundled into generators consumers and the transmission company Generators and consumers make contracts with each other directly and pay the transmission company to transport the electricity To take care of the inevitable errors a balancing market is run where generators can tender to produce extra power at short notice It would appear that the REDMan system is fundamentally compatible with this paradigm but further investigation is needed 10 5 9 Politics economics subsidies Many energy markets do not work according to pure free market principles Markets which have made the most successful transition to EG RE such as Germany Denmark etc are characterised by government intervention subsidising the new technologies The intention is of course to stimulate the market causing costs of the 152 new technology to drop through mass production until it is competitive even without subsidies This was successful in the case of wind energy in Europe Subsidies are already possible with the existing system to an extent because they do not affect the dispatching system directly Another mechanism which is sometimes seen is that some consumers are prepared to pay a premium price for environmentally sound renewable energy In order to allow this which is norrrational according to traditional economics the dispatching system could be modified to give source tracking of energy so that energy
145. erate ploy to make it more accessible to engineers and scientists To clarify Fig G 2 shows block diagram of the program The box containing the phrase Hello World is a string constant A small 99 piece of wire connects it to the right hand box labelled which is a terminal for a string indicator Clicking on the arrow in the top left hand corner causes the program to be compiled and run The execution can be thought of like this The Hello World string is emitted from the constant box and travels along the wire to the string indicator which then displays it on the front panel Fig G 3 As a finishing touch to the electronics metaphor the complete program is called a Virtual Instrument As the above description suggested LV is built on a dataflow model Unlike normal programming languages where the order of execution of different code segments is controlled by the order in which they appear in the listing in LV an object executes whenever data arrives at its input terminal The dataflow model is very intuitive and convenient for datalogging and analysis applications which involve moving quantities of data through successive processing operations However it can be bothersome in a few other situations e g when writing multithreaded programs 242 4 4 Datatypes The wires that carry data around a LabVIEW block diagram are roughly equivalent to variables or data s
146. erefore the real power cannot leave the faulted generator s and it they will begin to accelerate compared to the others in the system After a relatively short time a circuit breaker should operate to clear the fault The affected generators will be able to deliver real power again but because the faulted component is now out of service the amount they can deliver may be less Taking this into account the next step is to calculate whether the generators can dump the extra momentum they picked up while the fault was on If they can then the system 29 will settle into its new state and remain stable If the momentum is too great they will keep accelerating and pull out of synchronism The system is therefore unstable The margin of stability in a system depends on many things mainly the fault clearing time the inertia of generators the impedance and loading of transmission lines the demand and the topology of the network and protection systems Stability studies are used in a predictive capacity to plan the network so that it will deliver the most kW per unit capital cost with acceptable stability margins A popular definition of acceptable stability is that the system will remain stable if the single largest generating station or transmission line is lost 3 7 3 Application to EG systems This approach has been adequate for design of centralised power systems However a number of the assumptions will begin to fall down
147. erimental protocol and possible ways of interpreting the results This chapter describes the details of the experimental work and the results obtained 9 1 Testrun for energy balance transient Most of the data required for the experiments proposed could actually be gathered in the course of a single test run which exercised the various system components over a reasonable range A week in summer was chosen when a reasonable mixture of sunshine and cloudy periods could be expected The batteries were fully charged at the beginning of the experiment As regards the loading and the battery management setup there was some uncertainty Using the economic battery management presented some problems which were discussed previously To recap the issue was that the use of batteries was uneconomical compared to grid electricity so enabling the economic dispatch resulted in the batteries being turned off completely They could be artificially subsidised to make them work but there did not seem to be any way of doing this in a meaningful manner Therefore the economic algorithm was turned off and replaced by an ordinary cut out algorithm disabling the batteries once the state of charge dropped below 60 The inverter was set up with the revised economic dispatch algorithm A load of 50 watts was placed on the AC side and the buying price of this chosen so that it would be powered by renewable energy or grid power but battery power would be too expensive
148. ero crossing interrupt goto zero btfsc INTCON TOIF if a timer interrupt goto wave if a cosmic ray hits the interrupt circuitry we should cater for the possibility retfie e ck ck ck ck ck ck kk kk kk kk kk kx x zero crossing interrupt service routine 199 23 25 23 18 10 12 uses a simple digital PLL to keep locked into the 50Hz and reads the throttle setting protection also dealt with here ALS deal with the timer ASAP zero movf TMRO W get the current timer movwf tmrtmp movf errc W and restart it movwf TMRO coarse lock this adjusts timer speed until we get the correct number of steps per cycle movf index W bcf STATUS C sublw TARGET target index btfsc STATUS Z if the index is exactly target Z 1 goto fine do fine lock btfss STATUS C if the index is greater than target C 0 goto toohi execute too high too low executes if none of above conditions true incf errc F increment error goto fprot and skip over fine lock too high toohi decf errc F decrement error goto fprot and skip over fine lock once the coarse lock has worked the fine lock is brought into play TREF is subtracted from the measured TMRO value If TREF gt TMRO i e TREF TMRO 0 then the timer needs to be faster so we increment the error register Otherwise we decrement it
149. ers Since these can exceed 90 this could be challenging Safety and power quality were also crucial Any risk to personnel caused by malfunctioning of the inverter would be completely unacceptable as would disruption to other electrical equipment caused by interference from the inverter UK electricity companies drafted the G77 standard defining the required safety features and maximum level of distortion permissible for equipment connected to their grid and ideally the design would meet these However this is a fairly strict specification and there would always be the danger that meeting it would not be economical in terms of time and money at the time of writing the UK standard was harmonised with the US standard IEEE P929 2 The final consideration was the power rating of the apparatus It would be wise to make it at least the same if not more than the expected peak power from the RE sources in the experimental system The PV arrays totalled 260 W and there were also two 100 W wind turbines 156 4 Specification _ re eme CNN ULL emen _ e m Il eene mee 1 meme mue _ Reconnect time is to be reckoned as time after supply is restored within limits Disconnection to be by mechanical contacts to IEC 255 not electronic means A 5 Basic design choices Now
150. ery storage One reason is that while batteries might not be economical other storage systems may well be now or in future The use of storage is expected to increase as RE penetration increases hence REDMan like dispatchers would probably be involved in the control of storage systems and so it was thought important to demonstrate that it could be done Another reason was simply to absorb transient power fluctuations on the DC side and eliminate the need for MPP tracking which simplified the inverter design greatly Of course a bank of capacitors 147 would have done equally well in this respect but the storage capacity would have been insufficient for dispatching studies 10 3 Will it scale to large numbers The experimental system had two REDMan dispatchers one managing the AC domain the other the DC domain They co operated satisfactorily However it is still hard to extrapolate from this and say that a system of 50 100 5000 etc dispatchers will work properly too In fact some results gave slight cause for concern in that the coupling between the two domains seemed rather stronger than it should have been There was a kind of oscillation provoked by sudden changes of power levels in either domain However the data suggested that this was due to the inverter control algorithm which is effectively the component that interconnects the two domains interacting with the variable time stepping rather than any fundamental probl
151. esponsible As part of this task dummy load and generator programs were created The dispatcher communications and simulated loads generators were tested 125 together and seemed to be functioning correctly which is to say that they behaved according to the theoretical description of how the system should work First hardware The logical next step was to create some hardware for the software to control The first prototype dispatchable inverter was developed and also a rudimentary means of power measurement and load control called the Smart Socket These were interfaced to the computer by writing the first generator and load agents The inverter agent used was very basic treating the inverter as a dispatchable source of power at fixed price In this way the system could perform dispatching the generated power could be measured and seen to be equal to the demand within the accuracy limits of the hardware and when the demand was changed the supply changed to track it Mark Two hardware These results were encouraging so a decision was made to proceed to a more ambitious goal to demonstrate economic dispatching on a more realistic set of hardware similar to real world embedded generation plant Work was started on the most important component a more powerful efficient and reliable dispatchable inverter Once this was functioning a small solar generating plant was built using PV modules lead acid storage batteries
152. esults are shown in Fig 9 and Table 9 4 To summarise Efficiency appeared to have deteriorated by around 296 however this is within the limits of experimental error Also the measurement in this experiment included losses in the connecting leads between DC bus and inverter Calibration appeared to have drifted downwards by around 3 This cannot be explained by experimental error and must be a genuine problem with the inverter Temperature time dependence of calibration was checked The thermal effect noted in Appendix A was not detectable at this time 138 Efficiency vs power Efficiency 96 Error 95 Eff retest AC output W Fig 9 1 Efficiency and calibration curves of inverter re test Power Demand Expected DC Bus DC Input DC input AC Real Efficiency Power Power Setting power voltage current power power error error V A 0 0 0 0 0 0 0 5 12 27 5 25 1 39 34 75 20 9 60 14388 6 6 24 10 25 55 24 82 2 53 62 7946 48 1 76 59894 6 9 12 5455 15 38 82 5 24 68 3 64 89 8352 74 4 82 81832 8 1_ 9 81818 20 51 110 24 58 4 9 120 442 104 86 34862 6 5 45455 25 63 137 5 24 5 6 05 148 225 130 87 7045 7 5 5 45455 30 76 165 24 45 7 35 179 7075 158 87 92065 4 24242 35 89 192 5 24 38 8 6 209 668 187 89 18862 5 5 285714 40 102 220 24 29 10 1 245 329 216 88 04503 4 181818 45 114 247 5 24 25 11 1 269 175 241 89 53283 6 5 2 62626 50 127 275 24 2 12 5 302 5
153. et calculation was done to show the performance of the model over a discharge with parameters estimated as roughly representative of the batteries used in actual experiments Since battery operation is not economical compared to grid electricity the selling price had to be inflated to 50p kWh The buying price was 103 assumed a token 0 1p unit which was REDMan pricing for PV power at the time Figure 7 1 shows a graph of the estimated profit Since wear cost is quadratic with depth of discharge and income is linear the result is parabolic of the form bx ax Therefore two courses of action suggest themselves The greedy model will terminate the cycle at the maximum turning point of the parabola for maximum profit The altruistic model will terminate it where it intercepts the xaxis so as to just break even A compromise between the two could of course be arranged by terminating it at a target profit value after the turning point Of course these economic issues are specific to the particular storage system used Lead acid batteries just happen to be a very expensive way of storing energy once all the loss and wear mechanisms are taken into account Other systems such as flywheels regenerative fuel cells superconducting magnetic energy stores SMES and the like may well be much more favourable 7 1 13 Practical implementation The next step was to connect this model to the actual battery bank and the REDMan environment
154. etworks for example 5 6 Unfortunately these are designed with the expectation that they will run on computers in a national grid control centre 32 receiving data from the whole national grid and controlling all the generators It seems likely as argued in Section 2 7 that this approach would encounter serious difficulties due to the sheer number of units involved in EG Admittedly algorithms such as that described in 6 make steps in this direction by splitting the problem into several control areas but the method still appears to require substantial exchanges of data between areas and so does not seem directly applicable The ultimate goal in these terms could be described as the creation of an AGC algorithm incorporating LFC AVR and ED functions which is guaranteed stable with no data other than that which it can measure directly i e the voltage and current at the generator busbars and that irrespective of the number of instances of that algorithm deployed in the network and the reactances motor inertias etc also in the network A killer application like this would greatly enhance prospects for large scale embedded generation This is a serious task and one which may lead to economic compromises or in fact be impossible It would be very desirable to obtain some kind of theoretical argument that it is possible One line of attack might be to find a mathematical model of a worst case network comprising many i
155. evel could only decrease The solution to this was to periodically change the advertised AC power amount to full power If it turned out that there was an AC market for full power but DC conditions did not permit it the power would decrease on the next timestep However this represented one timestep s worth of DC power that was drawn without the dispatcher s permission and the more often this process was repeated the bigger would be the error In practice this was aggravated even further because different agents operated at different timesteps These issues will be discussed in chapter 10 The solution required rewriting the algorithm 7 2 6 Economic dispatch Mark Two The trick was to bid on the DC side for a slightly larger amount of power than the AC side was demanding If more DC power was available then the excess would be carried through to the AC side If there were a market for this extra AC then it would be bought The extra would be carried back to the DC side plus the slight increment In this way the power level is always driven upwards until availability or economic dispatching on either side limits it The drawback of this algorithm is that where the 112 set by user Selling price Available DC power HEDMan load YING price AC iclency available calculation DC Input estimate EN I Inverter control HEDMan power est generator From One DAQ timestep Battery
156. fault transient effects come in to play and the effect can be much more serious Transient analysis of networks is a whole other subject though and quite beyond the scope of this thesis 2 6 3 Economics Almost every type of generating plant is most efficient when operating at full power Part load efficiencies are often very poor Therefore dispatching is also concerned with maximising utilisation Put simply out of all the generators in operation as many as possible should be operated at maximum efficiency which normally means fully loaded This is in conflict with the requirement for reserve capacity and so a trade off has to be reached The exact nature of this trade off is a complicated issue because it sets the benefit due to quality of supply against the cost due to operating the reserve capacity In the UK with its proud tradition of a nationalised electricity industry where profit was less of an issue than quality of service the value of a dependable electricity supply is probably perceived to be quite high 2 6 4 Renewables When renewable energy is added to the supply pool it causes new problems The main problem is that the most popular forms of renewable energy hydro solar and 13 wind power are subject to the weather Therefore it is hard to predict exactly how much will be available This is really the same problem as coping with unexpected demands except in mirror image Wise design and siting of the
157. for research purposes This makes it attractive proposition to develop a simple ED algorithm specifically for the case in hand The first question to ask here is What is the goal of economic dispatch Politics notwithstanding it would be reasonable to say that it aims to provide the greatest benefit to the consumer at the lowest cost to the consumer So it would be sensible to put this objective at the heart of the prospective control system Then it could be hoped that the balance of efficiency stability and power quality appropriate to the situation after all it might change from place to place and from hour to hour will just fall out Of course creating an electrical network that optimises itself to deliver value for money sounds rather ridiculous But really it is just the same functionality as current ED systems stated from a higher level of abstraction If numbers can be derived from the real world electrical system that quantify value and money then there is at least a hope of using a computer to optimise value for money Now money is easy to define it is simply the unit cost of generating the energy But how to quantify value Capitalism states that any given item is worth what the market will bear From this it would be easy to define the value of a unit of energy as the maximum amount the user would be prepared to pay for it So defining value and money in these ways the problem can be s
158. forms of RE are expected to play an increasing part in the long term as fossil fuels become more expensive Price Wp market MWp 70 60 Module price 1994 const US Wp World marke a 40 o 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 Year Fig 2 1 Price trend of PV 1976 1994 reproduced from 12 10 2 5 1 Connecting EG to the grid The main problem lies in using EG alongside the existing electrical grid with its centralised generators While there is no fundamental reason why EG systems need to be connected to the grid at all there are advantages to doing so Firstly by connecting multiple generators together to form a network the reliability of the supply might be improved due to redundancy For the same reasons it is also very attractive to be able to combine the EG output with the existing grid There are also economic benefits the efficiency of a generator is usually a function of its power output and so it is usually possible to calculate an optimum power for each generator in a network in order to achieve the best overall efficiency In a similar manner networking will make CHP more efficient by making allowance for customers heat demands as well For example if a customer with an embedded CHP plant has a sudden large demand for heat he can increase the output of his CHP plant but then he will find himself with surplus electricit
159. from a simulation consisting of multiple domains or by extrapolating from 130 the settling time of two domain experimental rig It may well turn out that this is a very complex matter which cannot be resolved within the scope of this work 8 7 Objectives of experiments The following are suggested as realistic goals for the experimental program 1 Scan logfiles and check supplyzdemand also check transients 2 Measure actual power flows using separate instruments to ensure supply demand in real world also check transient behaviour 3 Test inverter performance and compare to model see 2 4 Check behaviour of agent programs 5 Try to quantify two domain interaction There will not necessarily be a one to one mapping between each of these objectives and a physical test procedure or experiment Experiments 2 and 3 are naturally related and the same physical experiment could cover both Similarly 1 4 and 5 could be covered by a single test run 8 8 Summary In this brief chapter a set of experiments were designed whose results would allow a judgement on whether the REDMan system was satisfactory and would hopefully suggest ways in which it might be improved The experiments were performed and data gathered Now it is time to discuss the results 131 132 Chapter9 Results of experiments The experimental apparatus was described in chapter 6 and the software it ran in chapters 5 and 7 Chapter 8 proposed an exp
160. g peaks The measured battery power shows the same peaks but also a drain of 5W during the hours of darkness This is not battery self discharge which cannot be measured by the DAQ system Rather it must be some load within the system Leakage in the wind turbines PV modules was eliminated it shows up zero in the measurements and also was checked with a separate meter Therefore either the dump load or the inverter must be responsible Stranger still though the battery then recharges slightly during the day even though the battery management agent is not buying any power the dispatched power trace shows zero 142 An explanation of this strange behaviour was sought through careful examination of all of the logged data In the end it turned out that the problem was nothing to do with the battery management program at all but with the inverter management which had two faults First was this When the available power is zero the economic dispatch calculation fails Instead of returning a buying price of zero for DC electricity as expected it returns Not A Number NaN The dispatcher appears to respond to NaN by treating it as though it were infinity and hence gives the inverter agent the go ahead to take power from the battery The inverter agent buys an initial 5 watts As soon as it gets power though the calculation starts to work again and the price changes from NaN to a low value The dispatcher refuses any further power and t
161. g it without having to go through intermediate communications links This will mean incorporating LFC into the grid connected EG storage units themselves probably as an algorithm running on a microcontroller chip The drawback to this approach is that separate EG units will not have knowledge of each others control actions and so there is always the possibility that they may act in an uncoordinated and unstable manner This possibility might be eliminated by proper design of the LFC algorithm Some kind of mult level scheme might also be desirable as described in 3 3 2 Similar considerations hold for automatic voltage regulation AVR functions 37 4 1 2 Embedding ED The essence of economic dispatch is co ordination between multiple generating units so that they share the load optimally Therefore it is meaningless to consider embedding ED algorithms in individual generating units ED requires some sort of control centre that connects to multiple generators However this work has already raised objections to the use of a single control centre for all generators belonging to a grid It seems more sensible to embed ED functionality at some intermediate level perhaps one control centre for every 10 to 50 EG units Modern computing technology would allow the control centre to be produced as a small box similar in size and cost to a computer network router and indeed it would probably connect to generators with a standard network technol
162. g the mains plug represents a very large disturbance which is easy to detect and triggers a complete inverter shutdown within 0 1 second before the plug has even been withdrawn far enough to allow contact with the exposed pins The loss of mains protection was tested extensively during development and in every case the test rig de energised immediately on loss of the grid For more details see appendix A The prospective AC loads were in a completely different part of the building with reliance made on the existing electrical wiring to transmit the power and on the building s Ethernet network to communicate dispatching information 6 4 5 Signal wiring The main part of this was between the current shunts and voltage divider on the distribution board and the computer s data acquisition card This was configured to give eight channels with differential inputs which were taken up by seven current shunts and one voltage divider Single ended inputs are impractical for use with current shunts because the common mode signal due to volt drops in wiring can be much larger than the voltage of interest developed across the shunt itself The variable dump load required a control voltage of approximately 1 5V to operate it over its full range of currents Since the DAQ card did not have any proper analogue outputs an R 2R ladder was used with its digital outputs to make a 89 rudimentary 8 bit DAC The dump load input was connected directl
163. hat has previously been validated by experiment For this project validation will be a serious problem The combination of power engineering power electronics computer networks and rule based decision making proposed here is to the author s knowledge a new one with no existing simulation tool immediately capable of modelling it So before simulations could be done a new simulation tool would have to be created and since it would be the first of its kind it would require validation by experiments So whatever direction the project takes in future there is a strong case for starting out with an experimental system 50 4 6 Designing an experiment The challenge here is to create the best possible experiment i e the one that gives most information towards reinforcing disproving the hypothesis subject to the limitations of money time and technical ability A system capable of demonstrating dispatching obviously requires more than one source of power and more than one demand for power It would also be desirable to incorporate storage systems and sources that are typical of those used in embedded generation There will of course be constraints on the amount of money and time available for the experiment Any practical system which could be built will probably be relatively small and biased towards more mainstream manageable technologies such as PV modules inverters wind turbines etc rather than mult megawatt CHP plants and
164. he inverter turns off again The cycle repeats as long as the available RE power is zero Inspection of the logfile measured powers shows that the inverter hardware does actually toggle on and off during the hours of darkness draining power from the battery So much for the unexplained discharging but what about the tendency to recharge during daytime The problem here lies with the inverter management lookup table Earlier when retesting the inverter hardware it was found that the power calibration had drifted downwards Therefore the inverter would have been outputting less AC power than the lookup table expected The efficiency also declined but by a very small amount only so on the whole the inverter would tend to draw less DC power than the lookup table predicted too So when the inverter management thought it was using all the available RE there would actually still be some energy entering the battery However this effect was not as great as the tendency to discharge during night time so overall the battery slowly discharged 143 Power W voltage V 200 150 100 a a 100 150 200 Voltage Power dispatched Power measured Time hh mm ss Fig 9 4 Time history of battery related variables 144 9 6 Two domain interaction The experimental system contains two instances of the dispatcher one for AC and one for DC The two are connected together by means of the in
165. ht have been to use ATP the Alternative Transients Program see 4 as the electrical simulator and to create a linkage so that ESP r provides the boundary conditions i e the amount of renewable energy being generated at any given instant E 7 References 1 Clarke J A Energy Simulation in Building Design Butterworth Heinemann London 2001 2 Kelly N J Towards a Design Environment for Building Integrated Energy Systems The integration of electrical power flow modelling with building simulation University of Strathclyde 1998 m Energy Systems Research Unit website University of Strathclyde http www esru strath ac uk 2002 4 Alternative Transients Program website Michigan Technical University http www ee mtu edu atp 2001 230 Appendix Hybrid PV System Controller This Appendix contains details of a control unit that was designed for a Hybrid PV project in Glasgow F 1 About the system Partick Housing Association commissioned this project to demonstrate commitment to the environment The roof of a tenement block was fitted with two arrays each of eight BP Saturn PV modules Each module gives 80 Wp and each array can supply up to 25A at 12V in full sun The arrays are fitted with a heat recovery system Fans circulate air over the back surface of the modules and the heated air is passed into the entrance hall and stairway of the tenement Electrical energy is stored in six 12V 75Ah lead acid
166. i DataSocket Read Boolean vi C Program Files National Instruments LabVIEW 6 vi lib platform dataskt 1lb DataSocket Read Boolean vi matchingl5 vi C My Documents labview backup labview labview phd demo matching15 vi rm global 0 vi C MMy Documents labview backup MlabviewNlabviewNphdNrealthingNrm global 0 vi 264 rchy Position in Hiera 265 2 5 REDMan dispatching algorithm matching15 vi matchingl5 vi Connector Pane Front Panel 10500202 20000 5000 266 Block Diagram there was not enough Fxceschdenend i poner Dou TEREE For each demand power available initialise shift reg auto index g B load dumped buy vector td inde pportunisticFlag size of source arra cost metric E demand is greater than or supply so buy the lot mm Sell vector Imake up jth demand from list of sources if cost too mt enable orl beL dumpload_ disable dumping Dump vector initialise buy vector jinitialise shift reg BB to get the source powers for next time Subtract the total buy vector from the list of source powers 267 Hidden parts of diagram peras prm 0 else p pd sump p iemin p pmax ci sumpe p_i cs sump p_i if there was only one sou
167. imate the ideal cycle more closely Unfortunately the law of diminishing returns holds true here too Adding one bottoming cycle might half the shortfall in efficiency but to achieve the full Second Law efficiency would require an infinite number of cycles Therefore the degree of improvement is again limited by economic and technical considerations 5 The one unchallenged assumption in all of this is that the heat rejected from a heat engine is wasted This is easy to overlook because in contemporary power plant design the heat is rejected at the very lowest temperature possible for efficiency s sake and this makes it of no practical use Deliberately raising the rejection temperature a little opens up a number of new applications for this waste heat for example space heating domestic hot water and industrial process steam This may seem like anathema because it makes the conversion to work less efficient but really it does not matter any more considering the rejected heat as a useful output the actual efficiency of the system becomes 85 90 4 The old efficiency now just determines the heat to power ratio This concept is known as cogeneration or combined heat and power CHP It promises to yield much greater gains in efficiency than any of the improvements to conventional heat mgines discussed previously Cogeneration schemes like these could obviously save a great deal of energy if they were widely used but this is
168. imesteps might be the ruling factor However this would presuppose the existence of another sub level of the system that could quickly supply or absorb the transient mismatches between commitments and real power flows If this sounds familiar it should be this is exactly the function performed by Load Frequency Control in conventional power systems see 3 2 or a balancing market as used in some trading systems Carrying this analogy further in a case where commitments are not strict the REDMan system itself corresponds to the economic dispatching 48 layer of a centralised system and like its centralised counterpart is no longer directly responsible for power quality See 4 1 1 4 3 2 4 4 4 Handling concurrency As described the REDMan system contains multiple entities any of which may act simultaneously The actions of some entities are dependent on the actions of others There must be some explicit means of scheduling the various actions in order to avoid total chaos and confusion An easy way of doing this is to link the actions to the system timesteps For instance at the beginning of the timestep a call would be sent out to sources and demands asking them to make their bids Once all bids had been received or after a time window allowing a reasonable time for bids to arrive the dispatching algorithm would be run and the results sent back out to the sources demands before the end of the timestep This is the same basic p
169. in EG systems The classical method reduces everything to real power transfer between voltage sources transient internal EMFs of synchronous generators through a network of reactances transmission lines In embedded generation the generators are often not synchronous but induction machines or even power electronic inverters Induction generators can pull out in a similar way to synchronous generators but with inverters the behaviour could be totally different Also EG units may be connected at the distribution level where it is no longer valid to view lines as simple inductive reactances see 3 3 1 above For this reason and other reasons the assumption of P Q decoupling 16 commonly made in power flow analyses can no longer be relied on Finally if distributed AVR frequency control systems see 3 3 3 were introduced these would have dynamics of their own which may spoil the usual assumptions of ideal AVR etc Of course modern power systems CAD packages can perform much more detailed analyses such as full transient analysis including the effects of user definable control algorithms and may well be able to manage this level of complexity However it will present a barrier to intuitive understanding for design engineers Also the computing power required to model a realistic network at the transient power electronic switching timescale and the effort required for model definition may be prohibitive This is only speculation
170. in Utility Interconnected Photovoltaic Systems Stevens J et al Sandia National Laboratories 2000 204 10 Switched Mode Power Supplies in Practice Kilgenstein O Wiley 1993 Electromagnetism for Electronic Engineers Carter R G Chapman amp Hall 1992 The Art of Electronics 2 ed Horowitz P Hill W Cambridge University Press 1989 Utility interactive photovoltaic power conditioning system with forward converter for domestic applications Matsui K et al IEE Proc Electr Power Appl Vol 147 No 3 May 2000 205 Appendix A voltage controlled dump load The experimental setup required an apparatus for getting rid of surplus electrical power In renewable energy this is known as a dump load In its simplest form it is a large resistor rather like an electric fire element which can be turned on or off by a relay as required This application called for something more advanced so that the power dumped could be modulated to any desired level It also seemed prudent to incorporate an overvoltage protection circuit which would dump power if the DC voltage ever rose to a potentially damaging level In normal operation of course the dispatching battery management system should keep the voltage within sensible limits but due to the experimental nature of things it might well not work according to plan B 1 Design considerations The design of the dump load unit was influe
171. in large networks A large grid will have many different generating plants of different kinds Some are capable of changing their power output quickly for LFC purposes some are more sluggish to respond and some for instance renewable energy plant cannot be controlled at all because they are determined by the amount of sun wind etc present at that time This limitation can be dealt with as follows All the generators are connected together therefore within limits imposed by the configuration of the network it does not matter physically where the injection of real power needed for frequency control comes from In fact the plant best suited to frequency control duties might change on a daily or hourly basis To make full use of this flexibility the modern trend is to schedule all generation on a minute by minute basis via commands sent from a grid control centre 4 This is done by advanced computer algorithms which combine load following control and economic dispatching The ED component calculates how the demand should be shared out amongst the available generators taking into account the reserves required for LFC Large networks are divided into areas each containing one or more generators plus some demand and the ED algorithm also calculates how much power should be exchanged between areas tie line power flow The LFC algorithm combines frequency error and tie line power flow error in a given area to give a quantity known as the Area
172. ines used approximately 100 source demand pairs could be dispatched and communicated with in a one second timestep The efficiency of this could doubtless be improved by orders of magnitude by using more advanced custom software but for the time being this speed is completely adequate for the size of the experimental system which is unlikely to contain more than 10 sources and 10 demands This means that DataSocket will be satisfactory and so it is time to define the data format and write the communications routines 66 5 7 _REDMan comms routines 5 7 1 Data items to be transferred The comms specification was defined in chapter 4 To recap a power source must state the maximum amount of power it can supply and the per unit price of this power In return it will be told the amount of power actually required of it A power demand must state the amount of power it wants and the price it is willing to pay In return it will be told whether it is authorised to draw that power In terms of data items then a power source sends two numbers and receives one number A power demand sends two numbers and receives one logical true false value Each of these items must be identified uniquely Furthermore when there are several sources and demands in a network the data streams associated with each client must be identified The primary means of identification in DataSocket is the URL and so it was a fairly obvious step to define a URL for e
173. ing the voltage at 2 4V cell for a few hours This is known as equalisation and ensures that all cells in a series connected battery receive a full charge even though some might have a slightly different capacity to others It also produces gas bubbles which ensure mixing of the electrolyte This should not be taken to excess or water loss will result 6 Electrolyte level should be checked from time to time e g every three months and any loss made good with de ionised water 7 1 10 Using these ideas How can these rules and methods of measurement be transformed into a computer program for managing batteries optimally First of all note that it is the act of discharging that wears the batteries out So if it was desired to manage the batteries for maximum life then they should never be discharged at all But batteries which cannot be discharged are obviously of no earthly use So the object of management cannot be to maximise the life Is it to maximise the amount of energy delivered from the store No because this means completely emptying the battery on a regular basis and so wearing it out prematurely The object of battery management should probably be to reconcile the conflicting goals of delivering the most energy and making the battery last as long as possible in other words to deliver the maximum benefit from the battery This can be defined as the ratio of how useful the battery is 99 expressed a price some
174. inverter Therefore the inverter buys more power The battery is full at this point so the surplus is directed to the dump load As PV generation falls again the dump load turns off at sample 31 Unfortunately the real story is slightly more complex The inverter output had already reached 50 watts at sample 9 why did the input continue to rise until sample 15 The answer is in the inverter control program which always tries to buy 5 watts more than it needs as discussed in Chapter 6 While the inverter is governed by the available input power it cannot find those extra 5 watts Once the output side demand becomes the limiting factor though the input power can be seen to rise 5 watts above the true value between sample 9 and sample 15 when the dump load turns on 140 Power Watts Power Watts 100 90 80 70 60 50 40 30 20 10 10 Dispatched powers on 03 09 02 12 02 49 to 12 09 49 PVPower Invinput DumpPower InvOutput Samples 10 sec Fig 9 2 Time history of dispatched powers Actual performance 03 09 02 12 02 49 12 09 49 PV Samples 10 sec Fig 9 3 Time history of actual powers 141 9 5 2 Actual dispatched power flows comparison The actual power flows measured by the DAQ subsystem can also be graphed in the same way This will give an idea of how the control programs for inverter
175. investigated ways of making such a system The objective of this project 1 to do just that but before producing any new ideas a more detailed review of power systems control is called for This is presented in the next chapter 2 9 References 1 BP Global Statistical Review of World Energy 2001 http www bp com 2 World Energy Prospects to 2020 International Energy Agency 2001 http www iea org 3 Nuclear Generating Units 1953 2001 Energy Information Agency USA 2003 http www eia doe gov 4 Cogeneration with Gas Engines Jenbacher Energy AG 2003 http www jenbacher com 16 10 11 12 13 14 Digest of United Kingdom Energy Statistics 2001 Department of Trade ad Industry http www dti gov uk p 168 Seven Year Statement National Grid plc 2002 http www nationalgrid com On line literature from Capstone Turbine Corporation 2002 http www capstone com Bowman Power Systems 2002 http www bowmanpower com U S Cogeneration Beyond reduction of pollution Cogeneration and On Site Power Production Vol 1 No 3 May June 2000 pp 25 33 Wind Energy The Facts Volume 2 Costs prices and values European Wind Energy Association 1997 p 56 Online at http www ewea org 2003 A hybrid PV wind energy module for integration in buildings Proc 16 European Solar Photovoltaic Energy Conference Glasgow May 2000
176. ion in the signal This was solved by having the opto isolator signal trigger a one shot multivibrator type circuit with a period longer than a half cycle of AC So much for the sensing side of things but there was also control to be done The simplest way of controlling the load was to turn it on and off with a relay built into the socket This was not a problem except that the specification called for information on the load s on off status even when it was turned off by the relay This 212 was achieved by fitting a small capacitor across the relay contacts allowing a small amount of current to pass not enough to operate the load but enough to allow the current sensor to tell if the load was turned on Unfortunately the capacitors chosen were slightly too big the result being that enough current passed to operate one portable radio tried during testing even when it was meant to be off There might have been a resonant condition between the capacitor and the magnetising inductance of the radio s transformer causing a greater current flow than expected C 2 Interfacing The prototype was constructed as a four way socket strip similar to an ordinary extension lead It had four digital inputs one controlling each relay via a transistor buffer and four digital outputs one from each current sensor These were all brought to a 24 line digital I O card fitted in a PC computer The card was a low cost model made by Maplin Electronics
177. ion is less effective 21 3 3 2 How AVR is set up In an ideal world the voltage would be regulated so closely that every consumer received exactly the nominal voltage This is not practical so compromises have to be made The voltages in the HV transmission system are regulated by reactive power compensation performed by generators and sometimes synchronous condensers or capacitors The aim here is as much to reduce losses by improving power factor as it is to keep the voltage constant Large generating stations may have an AVR system which controls the reactive power generation according to network voltage 12 or alternatively the generation of reactive power may be scheduled centrally Indeed a combination of the two is sometimes used where individual generators have their own AVR and LFC setpoints which in turn are controlled by central scheduling At the lower voltages used in the distribution system tap changing transformers are preferred There are so many of these that central control is not practical therefore they are normally controlled by their own local AVR which is set so that the voltage at the far end of the feeder never drops below the minimum acceptable The AVR may hold a constant voltage or it may droop to allow current sharing between paralleled transformers or it may even have compounding which makes the voltage rise as the current demand increases to compensate for drop in the feeder Finally in the furthest
178. ion will be upset On the other hand of course the simpler it is the less likely it is to be any use going down this path there is always the risk of blundering into a kind of electronic reductio ad absurdum where through disregarding vital information a nonsensical conclusion is arrived at So there must be a compromise between the two extremes getting a fine optimal solution at the cost of heavy computation and observing a large number of variables and on the other hand getting a solution which is quick and easy but useless The conditions are quite different to those in ordinary power networks the state variables of the system are much more difficult to observe there is more and faster control movement available and the concept of marginal cost may well be meaningless in many situations Therefore the compromise will naturally be different 4 3 Three simple rules for dispatching The compromise proposed is as follows by making a few sweeping assumptions the dispatcher s job may be reduced to a simple rule based algorithm This may be thought of as the knowledge elicitation phase where an experienced human operator is interviewed and his her knowledge turned into a rule base The most basic system of economic dispatching is well documented in the literature e g see 3 6 1 and hence there is no need to explicitly interview anyone as such As an absolute minimum the system can be expressed as three rules 4 3 1 The r
179. ipment is slightly more complex It is necessary for the power flow to from every item of plant to be known By known is meant that it should either be measured or controlled equal to a desired value depending on the type of plant For instance the power flow from PV modules to batteries can only be measured since it 1 controlled by the insolation whereas the power flow from batteries to dump load should be controlled It should also be noted in this context that when there are n flows of power into out of a node it is only necessary to know n 1 of them The nth can be inferred from conservation of energy which requires that the sum of powers must be zero assuming that negligible power is dissipated within the node itself which will be reasonable if it is chosen to be a small point Therefore in order to know the power balance for the DC side it is necessary to know the DC bus voltage and all but one of the currents drawn by equipment connected to it The AC bus has three power flows the power output of the inverter the power consumed by loads participating in the dispatching scheme and the power imported exported although export must be blocked for legal reasons through a notional grid connection that supplies the scheme This connection is not real of course because the loads participating in the scheme will be mixed in with other loads that do not and there will be no single connection between the portion of AC wiring serving pa
180. ircuit breaker which removes power completely from the component The goal of protection is to isolate faulty equipment as quickly as possible with minimal disruption to the supply of electricity and minimal impact on the stability see 3 7 below of the system 3 4 1 Types of protection equipment The most basic item of protection equipment is a fuse Fuses provide simple and reliable overcurrent protection with a choice of response characteristics However fuses are not suitable for many applications because they require replacement by hand before the power can be restored amongst other reasons Where this is a problem it is preferable to use a more complex system comprising protection relays and circuit breakers The protection relays are measurement devices which monitor current voltage frequency etc When the measured value is outwith the set limits the protection relay sends a signal to the circuit breaker causing it to open This system offers greater flexibility many different types of protection relays are available e g earth leakage phase imbalance differential current and several can be wired together to operate one circuit breaker Another very important feature is that the breaker can be programmed to reclose by itself Faults are often momentary in nature e g a lightning strike on overhead lines causing an arc to start In cases like these opening the breaker clears the fault and when it recloses the system resumes
181. ired is at least the square of the bit depth of the smaller of the two numbers Therefore each point will need at the very least 64 instructions say 100 So given that there ae 50 cycles per second and 4 quadrants in each cycle the processor needs to execute approximately 50 4 256 100 5 120 000 instructions per second Now to estimate the amount of memory required The LUT will contain around 256 values each of which will consume one word of memory Then there are 100 instructions executed to produce one data point This is not strictly accurate the program might consist of 10 instructions looped 10 times However it is adequate 175 a rough guess There will also be code for the phase locked loop and for reading in power commands This is assumed to be 100 instructions or less otherwise it would not have time to execute once per zero crossing Finally there will be perhaps another 200 instructions for setup and error detection Each instruction will also use one word so giving a total of 656 words Therefore a suitable microcontroller would have around 700 words of non volatile memory and an execution speed of about 5 million instructions per second MIPS Arizona Microchip s PIC16F84 10 chip looked quite attractive due to its simple reduced instruction set computer RISC architecture low price and ease of use it can be programmed using an ordinary PC and the very simple NOPPP No Parts PIC Progr
182. is fed through a transformer the transformer must be designed to operate at the inverter s output frequency In the case where the inverter operates at 50 Hz the transformer can be rather bulky and costly When a DC DC converter is used its operating frequency can be made different to the eventual output frequency By using a very high frequency such as 50 or 100 kHz much smaller transformer needed to handle the same amount of power making the finished apparatus lighter more compact and cheaper But since the inverter stage connects directly to the line in this circuit DC injection could be a problem When an inverter circuit is used to drive a transformer there are extra possibilities in terms of topology The most common circuit uses a centre tapped primary winding and cuts the number of switches required from four to two Inverter 1 in Fig A 3 is of this kind While this saves noney on switches it makes the transformer less efficient because each half of the winding is passing current only 50 of the time Hence for a given mean current which determines the power output the RMS current which determines the losses will be higher and so for a given design efficiency the transformer must be bigger and more expensive This must of course be weighed against the fact that in the full bridge circuit the current must pass through two switches in series In practice the centre tap circuit is very commonly used where the DC
183. is the electrical power flow simulation module which was added quite recently 2 E 1 About structure of ESP r ESP r runs on Unix and Linux operating systems with X Windows or similar GUI It is a large software suite consisting of more than a million lines of Fortran code The system is organised into a number of modules which are of three main kinds First are the problem definition modules which are used to create the model The output from these is a number of text files describing the model These files are next read into the simulator which solves the model and outputs a binary results database The simulator itself is a single program which performs an integrated solution of the thermal fluid flow and electrical domains It is composed of many modular subroutines which intercommunicate and access shared data by means of a Fortran Common block The module responsible for the electrical solution is the power flow simulator PFS Once the solution is finished the results database is read into the results analysis module RES which presents the results as graphs tables etc and allows export to other programs This greatly simplified description was correct as of Summer 2000 ESP r is under continuous development and the structure may well change See ESRU 3 for more details 225 E 2 The power flow simulator PFS PFS is an electrical power system simulator of the traditional kind The network is represented
184. items It will also obviously depend on the number of sources and demands with the following approximate relationships Time taken to bubble sort demand list varies as size of list Time taken to bubble sort source list varies as size of list Time taken for dispatching varies as size of demand list size of source list In other words the total run time of the algorithm is roughly proportional to the size of the problem squared One way to measure the actual speed of execution is by timing the algorithm as it is executed repeatedly with randomly generated source and demand lists In order to exercise the dispatching part of the algorithm to a maximum it is necessary to ensure that the demand list is always slightly smaller than the source list so that all demands must be dispatched and none can be dumped A simple program was written to perform this and tests conducted on a 300MHz Pentium II machine running Windows 98 SE and LabVIEW 6i 60 Execution time ms 2500 2000 1500 1000 500 y 8E 05x 0 0012 0 1634x 1 4764 1 50 100 150 200 250 300 Problem size number of generators loads Fig 5 3 Execution time of REDMan algorithm vs problem size 61 5 4 3 Speed test results Fig 5 3 shows how the execution time varied with the problem size The best fit to these results is a cubic polynomial which suggests that the problem is not O n as predicte
185. ive effects on system stability if a large amount of EG plant was present Therefore currently used protection hardware and software elements may 153 need to be updated which might be expensive and difficult since they are so numerous and widely distributed in the network These issues were discussed in detail in Section 3 4 As for what to do about them the first step is undoubtedly to perform system studies in order to gain a more detailed understanding of what is actually happening in the various interactions 10 6 Afinal word Many ambitious ideas have been put forward in this thesis and a basic proof of concept experiment has been performed However these ideas need more rigorous testing and refinement before they can form a system that will be of practical utility One very useful development will be to rewrite the REDMan dispatcher so as to take marginal cost into account This would resolve the problems and inconsistencies discovered during the experiments and increase its applicability to thermal generators inverters and the like However possibly the most important task will be system studies These will involve building a model of a sizeable power system that includes distributed generators and the REDMan control algorithms This could perhaps be done using simulation software like PSCad ATP or one of the advanced power systems simulators based on custom computer hardware The results of such studies will p
186. ken with other aspects of the control software In particular dispatching for the PV is a case of measuring the actual power fed into the battery by the PV and setting the maximum available power Pmax for dispatching purposes equal to this If the dispatcher decides that the whole of the PV power is not required then a dump load can be activated to draw the surplus power out of the battery As well as being a stabilising influence batteries are also an energy store and this aspect will need to be managed too In order to do this the battery will have to appear to the dispatching algorithm as two entities one source and one demand The 76 amounts of power bought and sold by the battery and the associated prices will depend on the state of charge of the battery in quite a complex way and this topic will be investigated in more detail in a subsequent chapter on battery management 6 2 3 Inverters Of course batteries and PV modules both work with direct current whereas the electrical grid is 240 volts AC 50 Hz In order to achieve co operation with the grid a piece of equipment is needed that will convert the DC to AC and change the voltage to the proper level This apparatus is known as an inverter In the system under consideration the inverter acts as a bridge between the part of the system containing generators and storage and the part containing the loads Therefore the power throughput of the inverter should be determin
187. l they became more robust but sacrificed intelligence and functionality According to this analogy it seemed probable that some elements of traditional power networks would need to be sacrificed But it seemed hopeful that nothing much would be lost in the sacrifice that much of the intelligence thrown away from the top might be reborn from the bottom up as an emergent property of the large complex system made from the interconnection of many small ones With this in mind an effort was made to abstract the automatic generation control dispatching process to the simplest set of rules possible without actually making it useless These rules were implemented in computer software as modules which could be distributed over networked machines The results were encouraging and motivated the construction of a small renewable energy generating plant on which to test the new control system This pilot plant was a success demonstrating desirable behaviour which would be impossible with existing EG control systems However there are still very many questions to be answered and issues to be explored before the REDMan system could see real world applications The work documented in this thesis is really just one drop and it can only be hoped that future years and the changing climates and priorities of the world will bring on the ocean that distributed generation needs Stephen Conner Glasgow November 2002 Chapter 2 Background This
188. lent me apparatus and generally did things that were beyond the call of duty Pat McGinness Eric Duncan and the Energy Systems technicians Donnie and Dino from DMEM Bill Rooney and Bob Baird from the Department of Electrical Engineering Alex Cairney from Dynamics and Control and Jim Docherty I was also fortunate in that when funds began to run out I found a part time job with an employer who understood the difficulties of juggling work and study and made sure that I was able to carry on with my research That job turned out to be very interesting and challenging in its own right So thank you Ivan Andonovic and everyone else at Kamelian Ltd for keeping my cash flow healthy and giving me my first U S Patent Last but not least I owe a debt of gratitude to my family who gave me great moral support throughout this whole business not to mention putting a roof over my head for some of the time lv Dedicated to the memory of my grandfather Ian Kitchener Colburn 1916 1972 Physicist and lover of knowledge ABSTRACT Electrical power systems are currently moving towards distributed generation using many small generators instead of a few large ones This can potentially produce great improvements in efficiency by allowing utilisation of waste heat cogeneration However it also poses new problems of control and co ordination of large numbers of generators which may be connected deep within the distribution network It is w
189. ler Sch controller Sch controller Sch Power supply and meter board pwr meter Sch Fig F 3 System wiring of the complete controller 237 Fig F 4 Circuit board for controller 238 Fig F 5 Circuit board for power meter module 239 Appendix How LabVIEW works The National Instruments LabVIEW system was used extensively for this work LabVIEW is not very well known outside the fields of data acquisition instrumentation and signal analysis Accordingly it was thought that a review of the capabilities and operating principles of LabVIEW would be of use to the reader if only to help in making sense of the LabVIEW block diagrams printed throughout this Work G 1 Whatis LabVIEW LabVIEW is a software program available for Windows MacOS and Linux operating systems intended for data acquisition and analysis tasks However it is based on a graphical programming model that allows it to be used for almost anything that other programming languages can do with the exception of object oriented programming It is a commercial product as opposed to opensource products G 2 Whatis it used for LabVIEW is mainly used in scientific and engineering applications A very popular use is for the automation of data acquisition analysis and testing procedures in research and industry In these applications it is usually used to operate and analyse data from data acquisition cards or GPIB enabled scientifi
190. line frequency This is fairly common in fact because power factor correction works in exactly this way If an inverter is islanded while supplying a load that matches its output with a strong enough resonance at 50 Hz the voltage and frequency will stay within limits and it will keep on going It was to address this situation that dedicated anti islanding protection was invented One popular antrislanding algorithm SFS SVS 6 is available in the public domain It functions by deliberately introducing instability if the voltage departs suddenly from its mean value the inverter is caused to change its power output in such a way as would amplify the disturbance A similar algorithm is used for frequency the inverter tries to change its own output frequency to amplify any deviations in grid frequency The thinking behind this approach is that the inverter is 194 basically made to behave like a hooligan and fight the grid instead of working with it Because the grid is bigger it always wins But as soon as an island develops the inverter wins the voltage and frequency go wildly out and the voltage frequency trips shut the inverter down Of course this approach assumes that inverters have an insignificant effect on the grid Therefore it is doomed to become a victim of its own success If enough inverters wth SFS SVS were installed in an area where the grid was weak they would actually be able to out fight it with tragi comic con
191. lists ensures that all subsequent transactions must also be uneconomical 92 This simple strategy is guaranteed to at least break even In practice it will normally make a profit the precise value of which is undefined A more advanced version of the algorithm could operate with a target profit margin specified in advance This would be implemented by keeping a running sum of the profit due to the fulfilment of each demand within the dispatching process profit 9 Pd Cd 9 2PCs ju i 1 Eq 5 1 where in a system with m demands and n sources is the power of the jth demand is the price it pays P is the power bought from the ith source and Cs is the price of that power The dispatching process would be terminated when this value dropped below a predefined profit margin However the simpler version will be considered for now Based on the description above a flowchart can be drawn for the dispatching algorithm Fig 5 1 In this the first step is to receive data for the sources the maximum dispatchable power and the per unit cost of it then receive data for the demands the price each is prepared to pay and the amount of power it requires Then the sources are sorted in order of the per unit price so that the cheapest will be used first and likewise with the demands so that the highest earning ones will receive priority The next step is an allocation procedure the dispatching proper Here the amount of power req
192. ll voltage however the real trouble comes when some of the smallest loads are used The same socket needs to be able to cope with loads like portable radios which might typically use 5 10VA 5VA represents a current of 21 mA hence a voltage across the same current shunt of only ImV This voltage is rather small and would require a precision op amp for detection This is undesirable because the amplifier besides costing money itself would require a power supply The power supply components would add extra cost and complexity to the system especially when requirements for safety isolation were taken into account A simple solution was to replace the current shunt resistor with a non linear element A string of diodes was ideal for this purpose the forward voltage of a diode is around 0 7V over a wide range of currents Three diodes in series gave enough voltage drop to light an opto isolator directly so the resulting circuit was very simple However it would dissipate considerably more power than allowed around 40 watts with a 13 amp load Although this did not meet the design goals it was not judged to be a problem in the research context The workaround consisted of limiting the current to 10 amps and fitting a metal heatsink to the socket housing Another problem was that the output was not continuous when current was flowing As the AC current passed through zero the opto isolator would momentarily go out causing an interrupt
193. lus power into a dump load where it is wasted as heat or by changing the loading on the PV module so that it is operating under a less efficient condition The former method is more simple because in this case the power incoming from the PV is always the maximum possible power which is therefore easy to measure for dispatching A dispatchable dump load was built for the purposes of this experiment The circuit and construction are described in Appendix B 6 2 2 Using batteries Batteries bring a major advantage in that they act as a short term energy store This means that the dispatching software does not have to keep the energy balance perfect at every instant in time as long as it adds up on average any transient errors will be soaked up by the battery To use an informal analogy they are a place for the energy to pile up while the dispatcher decides what to do with it This is a direct analogue to the LFC functions proposed earlier for distributed systems Obviously it is not quite the same since a DC system has no frequency But if voltage is taken as the stabilised parameter instead then the correspondence is exact the battery is an energy store that automatically lets out or takes in power to keep the DC control area stable It does this without any formal control algorithm as such merely by being a voltage source with low internal resistance Having such a flexible and forgiving kind of allows liberties to be ta
194. m the host computer via some kind of digital interface Synchronisation is achieved by sensing the zero crossings of the grid voltage and having the program adjust its timing until its own zero crossings coincide with them this is just a digital version of the phase locked loop used in electronics The choice between these two techniques is not difficult Although the analogue system is conceptually easier to understand there are serious challenges associated with making a VCO whose output amplitude remains constant to within 1 and a multiplier which is similarly stable The digital version while being somewhat more troublesome to construct and de bug avoids these problems altogether and so is naturally superior A 8 5 Choosing a processor The first task was then to choose a suitable microprocessor from the hundreds of types available In order to do this a quick estimate was made of the computing power required Firstly to meet the 5 distortion target 8 bit precision would be ample This then gives an idea of the number of points required per cycle if the output can only take 256 25 possible values then there is no point in sending out more than 256 points per quadrant of the sine wave Secondly for each point two 8 bit numbers are to be multiplied together Unless the processor has a hardware multiplier and only more complicated and powerful processors do this is quite an intensive task the number of instructions requ
195. m_load_open vi Labview labview phd realthing rm_load_close vi abview labview phd batman bmenergymeter vi abview labview phd realthing sysglobal vi abview labview p hd batman bmstatusglobal vi 251 tg oe B T Go JL Ed SS A EA LO X C ae 1 E LE e P A N m ier oP a H 2 2 BatMan state of charge estimator bmestimator vi bmestimator vi Connector Pane Reset 1 State of charge Mean power available w Sample time ms Buying cost Selling cost Estimated cost Front Panel 0 00 253 Block Diagram Cost of Cycle Completion estimate MAXDOD CB CBL 100 CD CB TTF INTDOD TTRDOD 2 1 1 CBUY CSUBJ NDM Q20 DOD Estimate time for a full recharge CBUY cu INTDOD power available Standing time to failure h DOD Deep cycles to m 020 20 Capacity ling osL peak detect Sample time ms gt 35000000 to hours 254 Peak Depth Of Discharge ntegrated DOD H 2 3 Smart inverter management invman7 vi invman7 vi Connector Pane Front Panel DC INPUT fw 200 300 400 500 00 500 Ao 0 00 AC OUTPUT 200 300 400 500 0 500 F 255 Block Diagram 0 0 DBL E Economic dispatching computes buying price based on selling price 3 output la
196. mperature Modelling these variations in voltage will be key to making an accurate estimate of the remaining usable energy at any time But here there is a serious problem because all of the relationships discussed above are subject to change as the battery ages 7 1 5 Ageing and service life There are many mechanisms at work here but this discussion will be limited to the two most likely to affect a battery in cyclic service The first is simply a reduction of the amount of active material lead sulphate takes up more room than metallic lead and so with each cycle of the battery some of the plate material is literally prised off by the growing sulphate crystals and falls to the bottom of the battery case where it is of no more use The amount of material shed in each cycle depends on the depth of 95 discharge Secondly if the battery is allowed to stand for a length of time while discharged the lead sulphate tends to change to a different crystal structure This new structure is very difficult to recharge and so areas of the plate that have become sulphated in this way are rendered useless The seriousness of this effect depends on the area of plate vulnerable to sulphation and on the length of time for which it is exposed Both of these mechanisms cause a reduction in the effective capacity and a corresponding increase in internal resistance To give an accurate estimate of available energy this must also be taken into accou
197. ms In any case all of the source files can be downloaded from ESRU s website at http www esru strath ac uk 5 3 Writing the program The first task is to define how the algorithm will interface with the other parts of the system The flowchart specified that the program should receive size and cost pairs for each source each demand send powers P to control hardware and 55 accept refuse jth demand In the actual program these items of information will be contained in data structures which are passed between the program and another program dealing with communications The usual way to deal with lists of this kind is to represent them by arrays of a custom datatype known as a typedef struct in C or a cluster in LabVIEW The cluster is rather like a database record in that it can be configured to contain any desired combination of numbers text Boolean true false values etc So the size and cost values for each source in the system can be represented as an array of clusters where the cluster contains a number representing size and a number representing cost and the array has one element for each source in the system Once the data are introduced to the program the next step is to sort them The indexed and sorted arrays are then fed to the source demand matching routine This operates exactly as shown in the flowchart The outputs from this are a list of the powers actually required from sources termed the bu
198. n and labour required for model definition might well be excessive So the most immediately obvious line of attack is to show by simulation or experiment that the system can control a smaller number of generators and then to prove by other means that if it can do this it will equally be able to control the larger number This will require finding rules and proofs of the inductive type pertaining to the behaviour of REDMan type systems These rules would be similar to shortcuts currently used in power systems simulation as discussed in Section 3 7 2 The REDMan system might for instance be tried on networks of different sizes with a view to imputing relationships between performance and size by which the performance on a really large network could be predicted or it might be tested on industry standard models like the two area three area multtarea model either in standard form or modified Doing this requires either an experimental power system or a simulation which can accommodate at least a small REDMan system But which is best to aim for Simulation is very attractive because it is a very cost effective way of doing things requiring no actual generators power lines and the like Unfortunately in the context of this project there is a major drawback a simulation cannot be trusted until it is validated Validation means checking the results of a test case against the experimental equivalent or against another simulator t
199. n for this it represents an error of only 10 parts per million and is hardly worth being concerned about unless of course it is an integraltype error that grows in an unbounded manner with time The next test is to compare the powers presented to the dispatching engine with the powers derived from system voltages and currents This will catch any internal inconsistencies in the agent programs This test is applicable only to the DC domain since AC voltages and currents were not logged The results are as follows all in watt hours generation is positive RE Generation 1393 559 1274 212 119 347 Battery net 72 516 165 967 93 451 Dump load 124 415 174 568 50 153 Inverter 1341 65 1265 611 76 039 Table 9 2 135 There are fairly large discrepancies here between the measured and dispatched values Both sets of values are internally consistent in that they add up to zero However this is no surprise since it was already shown that the dispatcher powers add up and the DAQ system works by taking n measurements and calculating the nth on the assumption that the sum of currents is zero The RE system generated less energy than the dispatcher thought to be the case This may have been due to an oversight in programming During development the wind turbine and PV currents were often observed to be a small negative amount in the absence of any wind sunshine This was assumed to be an offset error in the measurement electroni
200. n never be depended on to always give the same drain current In this application this can be dealt with by putting the dump load inside another feedback loop that measures the current and adjusts the gate voltage accordingly This could be a function of the dump load control software since the computer has access to the dump load current measurement The over voltage protection circuit would be easily made using a shunt regulator IC This is a chip that compares a voltage to an internal reference and if the voltage is higher than the reference sinks current to ground through a built in transistor A few of the popular TL431 regulator chips were to hand so they were a natural choice Unfortunately they operated in the wrong sense for this application an excess of voltage would require that the MOSFETs be turned on more to burn up more power In order to do this current should be sourced into the gate terminal but the regulator chip sinks current instead This would easily be dealt with by using a current mirror type circuit B 2 Circuit Fig 11 1 shows the circuit diagram 1 4 are the power MOSFETS sinking current to ground via resistors R1 R2 The gate terminals are connected to the control voltage input via R12 R13 and potentiometer R15 which adds an adjustable DC offset so that the MOSFETs are biased almost at their turn on threshold of 1 5 2V when the control voltage is zero R3 6 are stopper resistors to prevent parasitic
201. nced by the availability of parts as well as the required functionality A bank of power MOSFET transistors on a large heatsink was left over from a previous experiment These were almost ideally suited to the job except that the power dissipation was not quite high enough This could be improved by connecting power resistors in series with the MOSFETs There was still some doubt as to whether the heatsink could handle the power level though A few cooling fans would probably deal with this problem 200 watt wire wound resistors and 80mm 12 volt fans were available at reasonable prices MOSFETs are controlled by applying a voltage to the gate terminal They are normally used in the switching mode where the gate voltage is either zero or 10 15 volts enough to turn the device fully on By using intermediate gate voltages though the MOSFET can be turned only partly on In this linear region of operation the device looks like a current sink drawing a constant drain current proportional to the square of the gate voltage The constant of proportionality is quite poorly defined and would vary widely between different batches of devices so this mode of operation would not be very suited to a mass produced instrument For a one off construction though there is no problem the response of the actual devices can be measured and the circuit tailored accordingly It also varies with temperature 206 though so a particular gate voltage ca
202. nd also operated the relay according to the dump signal 7 5 3 Prototype power measuring agent This was the most desirable form of demand agent one that had accurate knowledge of the power being consumed by the load at the present time Getting this information is simple in theory by using a true RMS wattmeter which can communicate with a computer In practice it was difficult to find a meter like this and the few that could be found 10 were expensive and difficult to interface Eventually some information was found which suggested that one might be made by connecting current and voltage transformers to the line input of a computer soundcard and performing the appropriate signal processing in the computer The hardware and software are described in Appendix D This solved the problem and then creating the demand agent was easy a case of transferring the measured power value from the power meter program to the communications routine 118 set by user Amount Buying price REDMan load output to user Fig 7 8 block diagram of original demand agent set by user b REDMan load Fig 7 9 block diagram of smart socket agent system 119 current transformer from mains to load voltage transformer Soundcard Sampled data 11 kHz 16 bit stereo Signal processing Real power Computer Fig 7 10 block diagram of power measurement har
203. nd included as a subroutine in all REDMan components that required communications Within some of the connections defined by a single URL it is necessary to transport several data items This was done by means of attributes as mentioned earlier In the case of the connection from power source to dispatcher the maximum power vas given the attribute string and the cost Cs The connection in the opposite direction only carries one data item and so needs no attribute The connection from power demand to dispatcher carries two items the power required identified by the attribute and the price willing to pay In the reverse direction from dispatcher to demand there is only one item a logical true false informing the load if it is permitted to switch on 5 7 2 Allocating index numbers The previous section mentioned that the various agent programs are identified by index numbers The important question is how to manage these index numbers making sure that they are associated with the appropriate programs that they are really unique and that programs can join and leave the network if necessary Ideally the numbers would be dynamically allocated In other words an agent that had just started and wished to join the network would make contact with a separate management program to request an ID number The next free ID would be allocated When the agent terminated for example if its host applianc
204. nd the LED will be ON when the input exceeds the upper threshold set by VR1 and will go OFF when the input falls below the lower threshold The power on initial state will be OFF If R21 is installed instead the opposite will happen the output will be ON below the lower threshold and OFF above the upper one and the initial state will be ON Only one of the two resistors should be fitted F 5 2 Power supply metering module Refer to Fig F 2 This module supplies 5V to the other boards and also processes the meter point signals to drive two standard 0 10mA meters U1 is a standard 3 terminal regulator that reduces the 12V input to 5V The negative rail is generated by a miniature DC DC converter U3 This unit has a maximum input of 14V but under certain conditions the battery voltage may be more than this Q1 Q3 and 701 form a simple regulator circuit that limits the DC DC converter input to around 13 7V The output of the converter is nominally 12V and is regulated to 5V by U2 The meter amplifier section is based around U1 a quad op amp U5 provides a 2 5V reference which is buffered by to increase the current capability Likewise U1B buffers the measured value input from the voltage meter switch The meter connects between U1A and UIB outputs and so shows the difference between them This gives an expanded scale where 0 is 10V at the voltage controller input and 100 is 15V The current meter operates in a similar way ex
205. nduction motors In bang bang control the carrier wave 15 dispensed with and the modulating waveform is compared directly with the AC output The result of the comparison is used to control the power switching stage If the output is too great the power switch is turned OFF so that it begins to decrease and if it is too small the switch is turned back ON This can be thought of as forcing the inverter to generate its own carrier by self oscillating and for it to work efficiently there are two necessary conditions Firstly the comparator must have hysteresis a dead band where no action is taken and secondly there must be a low pass filter included between the inverter switches and the output which is being controlled If there were no filter the output would change immediately the power switch changed state Between them the hysteresis band size and the filter cutoff frequency determine the effective carrier frequency The bang bang approach is also wellsuited to grid intertied operation The conventional form of PWM generation is not very suitable because the modulating input controls the output voltage So the output of a classical PWM inverter appears 162 as a voltage source Now grid is also a voltage source and so there are two voltage sources connected together by the very small reactance of the inverter s output filter Thus tiny changes in the magnitude or phase of either voltage would cause large and dangerous curre
206. ngle generating plant but at a cost even though this spinning reserve does not produce any power it still uses fuel to make up its thermal and mechanical losses Centralised generation also means increased losses although the transmission system is very efficient it is not perfect For example 6 claims 1 7 losses for the UK National Grid transmission system However due to the enormous amounts of electricity being transmitted this represents a large loss in absolute terms in this case approximately 635 MW This figure varies according to network configuration and does not include distribution losses which are more difficult to measure but probably of a similar magnitude or even larger 2 3 1 The new idea embedded generation Fortunately there is a new way of organising the electrical network Instead of having a few large generating stations electricity generation can be subdivided into smaller units spread throughout the network This is known as distributed generation In the specific case where an effort is made to site generators as close as possible to centres of demand in such a way that buildings might almost be thought of as generating their own power it is termed embedded generation EG This approach promises to solve many of the problems discussed previously Firstly it makes CHP practical Individual generators can be sited right at centres of heat demand so there no problem with transmission CH
207. normal operation If the fault did not clear then the breaker will trip out again as soon as it recloses and after a few unsuccessful reclosing attempts it will normally trip out for good 23 3 4 2 Protection for EG Embedded generation has mostly the same protection systems as normal generating plant but with a few additional requirements The first issue is that EG units of small size using synchronous or induction generators cannot contribute much fault current in the case of a fault on the distribution network Sometimes the fault current is not much more than the full load current or even possibly less Nevertheless it could still hinder clearance of a fault and so the distribution network must be protected from it Overcurrent protection on the EG unit itself cannot do this though because the fault current flowing from the EG 1 too small In this scenario the fault will trip protection elsewhere in the system and isolate the section containing the fault and EG unit The EG will then hopefully trip out on over undervoltage or other protection 13 If it did not trip however a power island might be the result 3 4 3 Anti islanding protection If a part of the electrical network becomes separated from the rest due to a fault and it contains embedded generation it might continue to function independently Power companies believe this to be unacceptable for three reasons firstly it would endanger any personnel sent to re
208. ns routine is the power that must be dumped The conversion is simply a case of dividing by the measured battery voltage 7 4 2 Performance Reckoned in the steady state the dumped power was accurate to the limits of the DAQ system at least since an integral controller has no steady state error However the transient behaviour was a different story altogether The data acquisition routines update once every second and this represents a one second time delay in the integral control loop Classical control theory 9 predicts that unless the integrating time 115 set by user Selling price HEDMan generator Hardware Fig 7 6 block diagram of generator agent set by user Integral controller Dump Battery load voltage interface Dump load Actual dumped power current Fig 7 7 block diagram of dump load agent hardware 116 constant is made long enough the system will show underdamped behaviour and eventually instability Unsurprisingly experiments with the system showed this to be true A time constant of about 7 seconds was chosen to give an overdamped response The slow response was not a problem since the batteries were expected to make good any transient errors 7 5 Demand agent Throughout the development of the system a variety of demand agents were used were based on the same communications program except for the introduction of the Opportunistic flag The only real difference wa
209. nsidering that the University buildings like most office buildings all have built in Ethernet it becomes a very tempting choice Of course it is not perfect there are issues associated with latency and quality of service which may possibly not be good enough for this application These issues will be investigated in more detail in Section 5 7 5 below 5 6 2 DSTP DataSocket LabVIEW provides a variety of functions for network communication The most basic of these are the TCP IP VIs which as the name suggests allow the programmer to work at the low level of raw TCP IP data packets This is not optimal for the present application though it simply requires transmission of floating point numbers and logical true false values with the minimum of fuss To send these by raw TCP IP requires an intermediate stage of encoding floating point numbers into binary code This is not especially difficult but there are other higher level means of communication in LabVIEW which can make the job even easier DataSocket and the DSTP protocol DSTP is a part of National Instruments DataSocket system for network enabled measurement This is a unified way of dealing with scientific type data which allows it to be read from different sources in a flexible manner For instance a program using DataSocket could just as easily load a time series of numbers from a disk file on the local machine as from a remote file via FTP from a web site via HTTP or f
210. nstances of the algorithm under test along with reasonable models of tie lines and loads reduce this to a system of linear equations and apply stability criteria such as the well known Routh criteria from control engineering Since many power system components are non linear this may not be applicable Another possibility might be to test the algorithm in a classical two area or power system model using existing power systems modelling methods and perhaps developing new kinds of model boundary conditions that mimic a large number of similar adjoining areas without excessive computing requirements However each of these projects might well be a Ph D in its own right 17 is an excellent guide to the latest research in the area 3 8 Conclusions In this chapter the current trends in power systems control were reviewed with particular attention to techniques relevant to embedded generation The impression was that EG control was a relatively new and unexplored field where many of the traditional concepts did not directly apply The evidence suggested that 33 contemporary control and protection schemes used with EG units did not allow them to realise their full potential and might in fact interact destructively with existing systems causing negative effects on stability However there appeared to be no fundamental reason why this should be so on the contrary there was some evidence to suggest that EG systems might
211. nt 7 1 6 Calculating energy conte nt not easy With all these effects at work modelling a lead acid battery could well be a difficult job Nevertheless it has been done in considerable detail 2 although models for ageing are less common The main objection to this approach is the amount of experimental effort required to characterise the battery in sufficient detail To gather data on capacity variation with discharge rate multiple charge and discharge cycles at various different rates must be done which takes a lot of time and generates a lot of data to be processed More worryingly the mere act of cycling the battery ages it and so reduces its capacity over the duration of the experiment the battery you finish the series of experiments with is not the same one you had at the start Gathering data on ageing effects is more time consuming still and involves the destruction of at least two batteries And even if the experiments could be done there were grounds for suspecting that a fully detailed model would not be very robust merely because of its complexity The difficulties of making a full model seemed very discouraging So it was decided to take a more basic approach and make do with the simplest model that would give reasonable results 7 1 7 A simpler battery management model To assist in making this new model battery management practices currently used in industry were reviewed The information presented in 7 1 8 and 7 1
212. nt surges However by sensing the inverter s instantaneous output current and using bang bang control acting on this then the inverter appears as a current source instead and the problem is avoided The necessary low pass filter takes the form of an inductor in series with the inverter s output terminals A 6 Simulation An inverter of the design described above was simulated using ATP the Alternative Transients Program 5 At this early stage the precise design details were not yet known For this reason and also to save computing resources a simple representative circuit was simulated The model used was a combination of two parts a schematic of the switching circuit created with the ATPDraw graphical front end and a description of a bang bang controller algorithm circuit made with ATP s MODELS language The ATP simulator solved the electrical circuit and also executed the controller algorithm at every timestep of the solution This model was mainly used as a design aid to explore the effects of different kinds of output filter and different hysteresis bands The results gave confidence that the proposed design could be made to work and give good performance Figs A 4 A 7 show the schematic diagram the control algorithm description and some sample waveforms The circuit modelled here uses a second order LC output filter A 7 Practical design issues Once the basic topology and strategy of control had been developed the nex
213. ntally the same At first sight this seems to be a simple matter It costs money to generate electricity and electricity from different sources is interchangeable thanks to a national grid So it should simply be a case of ranking generators in order of cost and using the cheapest sources first Of course in reality there are a few catches First the price of electricity from a given source is not constant but rather it depends on the amount bought In the short term this is because the efficiency of a generating plant depends on the loading Engines turbines boilers etc work most efficiently at a certain loading which is determined at the design stage Any other operating point compromises the efficiency Hence more fuel which costs money is wasted and the cost of the electricity must be raised if the plant is to stay in profit The cost calculated on this basis is called the short run marginal cost Also in the long term the capital cost of the plant plus interest on loans must be repaid out of the earnings and this component will also loom larger if the plant is operated at a low capacity factor Including this component gives the long run marginal cost Finally some of the generated power is lost in transmission therefore the marginal cost also should reflect this The amount depends on the distance between source and demand and the characteristics of the transmission system In practice this complicates matters grea
214. ntral star point With respect to the AC side matters are more complicated since loads sometimes have nor unity power factor or indeed may draw a non sinusoidal current Accurate measurement under these conditions requires a dedicated power meter circuit These generally function by rapidly sampling the instantaneous voltage and current and using digital signal processing techniques to multiply the voltage and current samples together and take a running average of the result which will be true RMS power 2 However as previously discussed some loads have a constant and predictable power consumption and these can be measured by just sensing whether they are ON or OFF To sum up there are three possible levels of measurement depending on the nature of the load shown here in order of difficulty and expense 1 An ON OFF indication for loads whose consumption is constant and whose power factor is constant although it need not be unity 2 Corrected average rectified current for loads with unity or constant power factor but whose consumption may vary 3 True RMS wattmeter for loads with neither constant power factor nor constant consumption 80 In the experimental system the ON OFF sensor type 1 is attractive on grounds of being simple A sensor of this kind was built and tested this is described in Appendix C A true RMS wattmeter with computer interface was also built see Appendix D The responsi
215. nts of 5 or 10 tolerance electronic level control circuits whose characteristics can vary with temperature and electrolytic capacitors whose value can vary by up to 80 Problems which might be caused include calibration shifts with temperature unwanted phase shifts and mismatches in phase shift between the two stereo channels Any of these could degrade the accuracy of a power measurement and unfortunately it would be difficult to quantify the accuracy of a particular soundcard since it would involve testing it over various different temperatures and so on It seemed that a reasonable plan of attack might be to build a prototype of the system and see if it worked D 2 Hardware The most important parts of the system apart from the soundcard would be the current and voltage transducers The requirements for these would be galvanic isolation between the mains voltage and the soundcard and reasonable accuracy in reproducing the voltage current waveform the latter implying a good bandwidth free from phase and amplitude distortion etc This requirement might be quantified as total harmonic distortion less than a few percent and bandwidth of around 5 10 KHz The simplest kind of transducer meeting these requirements is a current or voltage transformer A transformer has one drawback in that it cannot reproduce any DC component in the waveform This should not be a problem though because AC supplies and loads are not supposed to have
216. nverter requires a filter inductance in series with the load Quite a small inductor had been used wit the intention of using the leakage inductance of the transformer to help with the filtering Unfortunately the transformer winding also has capacitance to the core which is earthed In use one end of the winding was connected directly to the unfiltered HF output and the very high rate of change of voltage dV dt caused large transient currents to flow to earth via the winding core capacitance Swapping the leads around so that the inductor was in series with the end of the winding having most capacitance to the core reduced the problem considerably A more permanent fix would be to use two inductors one for each winding end 10 2 Excessive losses The Mark One also suffered from excessive losses It was only 85 efficient at rated output No single component was really responsible the losses were equally spread amongst the transformer windings transistors and current sense resistors A 10 3 Too much distortion Also it did not meet the 5 distortion target The source of the distortion turned out to be the transformer a commercial unit designed with economy in mind Discussions with a manufacturer of transformers revealed that it is common practice to design for a peak flux density of 1 5 Tesla which is actually greater than the saturation point of the core material 1 3 T Therefore the core saturates towards the 180
217. ny of this However there did seem to be one or two small areas that might be susceptible to technological advances Distributed generation was a recent fairly radical concept in which large central electric power stations are replaced by large numbers of small generators If these are sited carefully where heat is also demanded perhaps in each domestic dwelling or public building the waste heat inherent in electricity generation need not go to waste With further study it became apparent that this concept was missing one important component a way of controlling all the small generators synchronising them together so that they worked in harmony keeping all the advantages of the old electrical grid No record could be found in the public domain of any work towards this objective and so it seemed like a promising avenue for research The first step was to study the automatic generation control systems used in ordinary power networks The tendency in these was for generation to be scheduled centrally and it was obvious that this would never work with distributed generation where a network might contain millions of units If it was to be successful the control system would have to be distributed too There are not many precedents for control systems of this kind A rough analogy can be drawn with biological systems e g plants and rudimentary animal life forms These all showed the same tradeoff by forfeiting a centralised contro
218. o EG Systems oe ee eite det d ete eee tens 30 3 74 How will EG affect siabilify2i a e e ee eee dete 31 3 7 9 ied e e eite edet ees 31 3 50 OPperhaps 32 3 87 GONGCEUSIONS aec im avete te a E OR 33 3 9 REFERENGES 5st cem tete ed nep ee ee de eate e Hed tee 34 CHAPTER 4 4 NEW DISPATCHING SYSTEM PROPOSED 1 5 e eeeete tata tantus sata tenen enn 37 4 1 DISTRIBUTED DISPATCHING 37 4 1 1 Embedding LFC and AVR eerte tert tenete Eee 37 4 1 2 Embedding ED c RR ARA S 36 4 1 3 Embedding demand prediction 36 422 ECONOMIC CONSIDERATION Ssa RR e RARE S SERPENS 38 4 3 THREE SIMPLE RULES FOR DISPATCHING cccscsscssssesscsessesecsesesseseesessesessessesesscsceaeseescsacaseecsacaceecaeees 40 4 5 1 E 40 4 3 2 Assumptions underlying the rules eese tette nennen 41 44 REDMAN PROPOSED tiene etie ned ee teet i ch dee s 43 4 4 1 Algorithm and communications 43 4 4 2 Timestepping CONSIGCTALIONS s secesesesesesesereseseseseseseseseseseseseseseseseseseseseseseseseseseseseseaeseseaeaeaeaeaeaeaees 46 4 4 3 Timestepping and commi tment eese estet tete tenent netten netten nente nnn 46 4 44 Handling concurrent issan e Sede er e e R 49 45 PROVING THE CONCEP Ds certnm ere TOT X a te or ees 4 6 DESIGNING AN EXPERIMEN
219. ogy such as Ethernet DSL ISDN etc It also seems logical that this should be done on a hierarchical basis In other words the combination of a control centre and its generators could be made to appear as if it was a single larger embedded generator which would connect along with other similar units to a higher level ED control centre and so on until the entire EG fleet was reduced to a few top level controllers These might well be functionally similar to large generating stations and capable of integration with the existing AGC systems for centralised generating plant 4 1 3 Embedding demand prediction There is scope for demand prediction at all levels In tune with the embedded dispatching philosophy though it seems logical to embed it in those components that can benefit from it Storage units are the most obvious example although there are many electrical loads which could benefit from being able to predict the most economical times to operate 4 2 Economic considerations As was seen previously there are advanced algorithms which attempt to operate all the generators in a power network at the economic optimum point It is quite possible that existing ED algorithms could be modified so as to only optimise their local subnet of generators but they seem to contain much functionality that is not applicable to the case in hand Furthermore those existing ED codes which could be 38 found are commercial and not available
220. oint the way forward in development of the algorithms and hopefully hasten the day when REDMan can be let loose on a real power system 154 Appendix A Building an inverter A 1 Whatis an inverter Electrical power is usually transmitted and used in the form of alternating current However some kinds of electrical generation and storage devices produce direct current examples being PV modules and batteries An inverter is a power electronic apparatus which converts DC to AC allowing the DC power from these generators to be used with ordinary AC appliances and or mixed with the existing electrical grid A 2 Why build an inverter During the course of this project the need was identified for a novel type of inverter which could be dispatched as part of the REDMan system In practice this meant that it should have a computer interface that could accept commands to set the output power No inverter of this kind was available on the market and it was not even remotely economical to have one custom made by an outside contractor So there seemed to be two possible courses of action buying a commercial inverter that lacked the required facilities and modifying it to suit or building an inverter from scratch Modifying a commercial unit seemed attractive at first but after a few inquiries 1 it became apparent that manufacturers kept the details of their inverters confidential They were not prepared to release any information on the
221. oping with the power quality consequences of failure These means are already used in centralised power systems and were discussed in Section 4 4 3 151 10 5 7 Combined heat and power dispatching More research is required into the simultaneous dispatching of heat and power flows of CHP generators There would be two possibilities in this respect The most obvious is to use two dispatchers for each domain one for the electricity market and another for the heat However this does not explicitly manage the interactions between the two markets A more logical line of attack would be to expand the dispatcher algorithm to deal with several inter related energy flows associated with each source rather than a single stream per source as at present Solution of the heat and electricity markets would then be simultaneous and the interactions would not be an issue practical implementation of this would involve revising the dispatch algorithm and communications layer as described in 10 5 3 above In addition to these changes though modelling equations for generators would include means of calculating heat output as well as power output and demands would indicate whether it was heat or power that they required 10 5 8 New Electricity Trading Arrangements Around the time that this project was in progress many electricity utilities including the UK s were transitioning from the old pool trading system to a new paradigm based on bilat
222. ower systems performance if it was scheduled in a suitable manner That is a very large and so it is only fair to present the other side of the case There are two main problems The first of these is renewable energy which will probably form a significant part of the EG mix in future years In 3 7 5 it was assumed that the power output of EG units could be controlled entirely at will This is true for such technologies as fuel cells storage units of all kinds and combustion engines turbines whether fossil or biomass fuelled These all have an inherent store of energy that can be released as necessary However some RE generators like photovoltaics and wind turbines are not directly dispatchable The only possible means of control is dumping of any excess output also known as constraint management But on the whole the output is controlled by the incident solar or wind energy and so is liable to fluctuate more or less at random This could obviously have a negative effect on power quality In practice other EG units that are dispatchable could be used to counteract the effect This possibility has already been investigated for instance in small scale wind diesel installations The other problem which has already been discussed here is that the algorithms for control of large numbers of embedded generators do not exist The only precedents are the automatic generation control AGC algorithms used in conventional centralised n
223. pacity variation with discharge rate 7 1 9 Taking care of batteries Lead acid batteries have been in use for over 100 years and in that time a great amount of empirical experience has been gained in how to treat them for optimal performance Today this body of knowledge is summed up by a few simple rules which will be reviewed here These guidelines apply to batteries in cyclic service as in renewable energy systems 1 Lead acid batteries do not like to be discharged more than absolutely necessary and should certainly never be discharged completely The exact amount of discharge tolerable depends on the battery design purpose built deep cycle 98 batteries can withstand perhaps 2 000 cycles to 20 capacity 1 80 of charge extracted standby and automotive types perhaps 500 cycles to 7096 capacity 7 2 Lead acid batteries should not be allowed to stand discharged any longer than necessary battery which is left completely discharged will be destroyed in a few months 3 The maximum allowed charging current depends on the state of charge it is very high when the battery is almost empty and tapers off as it charges This requirement is usually met by a constant voltage charger with current limiting 4 Charging is considered complete when the terminal voltage reaches 2 4 volts per cell i e 14 4 V for a nominal 12 volt battery 5 The battery should be overcharged occasionally e g once a month by maintain
224. pair the fault secondly the island would drift out of phase with the rest of the network and it would be impossible to reconnect it without causing a catastrophic transient and thirdly the power quality in the island could not be guaranteed On these grounds they claim that the only safe way of managing islanding is to forbid it islands must be made deliberately unstable so that they trip themselves out within a few seconds Unfortunately these antrislanding methods are exactly the opposite of what 15 required for good network stability For instance if the frequency begins to collapse the grid needs a big injection of real power as quickly as possible Instead the rate of change of frequency ROCOF protection systems commonly used with EG are likely to trip out the embedded generators which will take more real power away and simply aggravate the collapse A similar objection can be levelled at the SFS SVS anti islanding algorithms for power electronic inverters 14 Heavy penetration of EG systems with this kind of antrislanding protection would probably have disastrous effects on network stability So it seems that if EG is to achieve 24 significant penetration the current philosophy on islanding may have to change completely The first objection that islanding is unsafe is perhaps not insurmountable Electrical engineers are all trained in safety and always assume a circuit to be live unless they have first hand evidence
225. parameters that are directly measurable In electrical terms energy is the product of voltage current and time Now the product of current and time i e charge is the basic measure of electrochemical reactions The rules of chemistry dictate that X coulombs of charge will always cause the reaction of Y moles of substance Hence the mass of reactants available in a battery sets a definite upper limit on the amount of charge that can be stored in it This is why batteries are specified in terms of amp hours and this is the accepted definition of capacity the number of amp hours available from a battery 7 1 3 Variation of capacity with discharge rate The catch is that the available mass of reactants is not constant In order to maximise the surface area of material that can participate and hence the capacity battery plates are made porous rather like a sponge made of lead As the battery discharges the surface of the plates is covered in lead sulphate At low rates of discharge it builds up evenly and the reaction proceeds until most of the lead has been converted But 93 at high rates of discharge the resistivity of the electrolyte in the pores comes into play and causes current to flow mostly from the outside surface of the plate More current means more lead sulphate created and so the sulphate builds up mostly on the surface blocking the pores so that the lead inside cannot react The net effect is that the c
226. peatable see Appendix A 13 also Section 9 4 7 2 4 Simple fixed price algorithms The first control algorithm tested had both buy and sell prices fixed The inverter power level was set according to demand on the AC side If this demand plus losses could not be met by the DC side at the specified price the inverter was dumped in the same way as an ordinary load 111 7 2 5 Economic dispatch Mark One In the original economic dispatch algorithm the selling price was fixed and the buying price controlled It operated in two phases In the first phase it attempted to buy power from the DC side with the Opportunistic flag set and a maximum amount equal to the largest input the inverter could take The price was set equal to the selling price plus the amount required to cover losses The amount of power actually returned by the dispatcher was then advertised for sale on the AC side less losses Once the dispatcher returned the desired AC power the second phase of operation began The inverter was started and the initial power and price on the DC side were replaced by the measured DC power and the price calculated from the actual losses There were some problems with this approach Firstly it had somewhat of a one way characteristic Once the second phase was underway the advertised power on the AC side was set by the available DC power and the maximum DC power was in turn set by the amount of AC power bought so the power l
227. perform just as well as centralised ones if each EG unit was controlled with the kind of care and detail devoted to AGC of centralised generators at present Extending these AGC techniques to large numbers of embedded generators was identified as a very promising area for further work Unfortunately though it seemed that the existing AGC algorithms might not be directly scalable and fundamentally new techniques would be required In the next chapter the question will be explored of what those new techniques might be 3 9 References 1 Feedback Control of Dynamic Systems Franklin F G et al Addison Wesley 1994 2 Embedded Generation Jenkins et al IEE London 2000 3 ibid p 258 4 Power plant control and instrumentation The control of boilers and HRSG systems Lindsley D IEE London 2000 5 A new approach to automatic generation control and economic dispatch for reat time closed loop operation Albrecht J et al Proc 12 Power Systems Computation Conference Dresden 1996 6 Decentralized Load Frequency based on H Control Ishii T et al Electrical Engineering In Japan Vol 136 No 3 2001 7 Newnes Electrical Pocket Book 18 Edition ed Reeves E A Butterworth London 1981 8 Higher Electrical Engineering Shepherd J et al Pitman London 1970 p 504 34 10 11 12 13 14 15 16 17 18 19 20 21 ibi
228. performed reliably and without errors such errors as were present could be attributed to problems with the hardware or auxiliary agent programs These errors were an inconvenience but did not detract from the proof of concept 10 2 Does it work economically Within the assumptions made zero marginal cost etc it certainly does The question is rather one of whether those assumptions were valid in the first place The inverter actually did not have zero marginal cost and so the dispatcher had no formal means of dealing with this The inverter management program had to find a way of turning the incremental cost into a single figure In practice this was done by calculating the price as a function of the power level at each timestep and just presenting this figure to the dispatcher This worked properly provided that only one device in each dispatching domain operated in this way Also there were problems with the economic battery management which had to be disabled Partly it was a victim of its own success the problem being that the lead acid batteries themselves were not economically viable for use in this system It also suffered from a minor bug caus ing some power to be wasted Of course this begs the question If the batteries were not economically viable why were they included in the system in the first place It would have been possible to have an inverter that connected directly to the PV modules with no intervening batt
229. power sink of the kind required to model the inverter input Using it as such is just a matter of accounting that is picking up the imported exported power value of the node excluding it from the normal post processing calculations so that it does not look as if the power is leaving the network instead diverting it to the input of the inverter model The only disadvantage to this method is that the inverter input voltage is fixed hence it stays the same at all times This is not an accurate model of a real inverter where the voltage must vary for MPP tracking Hence it precludes use with an accurate PV model This was not a problem in the initial tests because the existing PV model is idealised with a built in MPP tracker It does not really care about the voltage of its output node and always outputs the maximum power regardless So the inverter input node could be fixed at a nominal input voltage selected by the user The end node of the inverter on the other hand is just a power source This is implemented using the same routine as for a load but with the sign reversed Instead of the load getting its magnitude by the ordinary method i e from a model description input by the user or from the building plant side in the case of a hybrid component it is set by the output of the inverter model To summarise then the inverter model is not really part of the power flow network The input and output nodes are boundary node
230. r suspecting that the true cause was not heating 190 Current amp Fig 60 50 DC Bus current amp BR o 10 Time sec 19 Mark Two inverter line current at 550W output Inset shows waveform at current peak magnified to show 200 us 0 0 005 0 01 0 015 0 02 0 025 0 03 0 035 0 04 0 045 0 05 Time sec Fig A 20 Mark Two inverter DC input current also at 550W output 191 Efficiency Error Q Var AC output W Fig A 21 Mark Two efficiency reactive power control error 192 A 14 Lessons learnt from the Mark 2 The lack of efficiency was the most intractable problem Again from an examination of the temperatures of the various components while running under load it seemed that no one part was contributing excessively to the losses There were probably two main causes an underestimation of resistance across the whole circuit by not taking into account the skin effect and an ignorance of iron losses in the transformer which would also be aggravated by the high frequency currents and any DC component in the output The premature power limiting was also puzzling Considerable time was spent in investigation of this without any real success In the end the most likely explanation was transient dips in the power supply rail caused by circuit inductances and the extremely high rate of rise of current up to 700A s in the rail
231. r technology and the efficiency improvements promised by cogeneration Small medium heat engines suitable for cogeneration have seen many advances Pollution reliability and noise control have been improved by borrowing technologies from automobile and aero engine design Microprocessor control allows generators to operate automatically without supervision High temperature ceramics and alloys together with computer aided design have increased the efficiency of smaller heat engines Examples of these modern heat engine technologies are microturbines 7 8 and IC engines 4 These are popular for cogeneration or backup power in the industrial and commercial sectors 9 Renewable energy has seen improvements too For instance the cost of photovoltaic modules has fallen steadily as shown in fig 2 1 Wind turbines have also become more cost effective although it is harder to quantify prices in this case because of the variable nature of output 10 Some renewable generators have also been designed specifically for embedded generation for example ducted wind turbines which are an integral part of the building fagade 11 However these are currently at the experimental stage At present renewable electricity is mostly uneconomic compared to fossil fuel generated electricity hence such large PV and wind installations as there are mainly exist because of government subsidies However wind power is rapidly becoming competitive and other
232. rce True P 268 List of SubVIs SOR sort vi ME Documents labview backup MlabviewMlabviewNphdNdemo Nsort vi SOR Sort dem2 vi REV C MMy Documents labview backup labview labview phd demo sort dem2 vi 269 270
233. re multiplied together to create a third array the time series of instantaneous power The mean of this series was taken giving the real power P Second the wariances standard deviation of both voltage and current series were calculated These were the RMS voltage and current respectively The product of RMS voltage and current gave the apparent power S Third the reactive power Q and power factor P F were calculated by the well known relations P 4Q S P F P S This completed the power analysis D 4 Calibration and testing The first step in calibration was to set the voltage reading A suitable voltage calibration constant was chosen in this case 350V f s and the voltage trimpot was adjusted so that the power analyser voltage readout agreed with a voltmeter connected to the power analyser socket Next a current calibration constant was chosen in this case 8 amps a load of known power consumption was connected and the current trimpot adjusted until the power readout agreed The choice of calibration constants is not arbitrary rather it determines the full scale range of the device In this case it would 1400VA That assumes BE sinusoidal voltages and currents in practice some extra headroom must be allowed crest factor 220 The calibration was retested at regular intervals and was found to drift slightly The same 60 watt bulb read between 60 and 62 watts at different times Of course this
234. renewable system can help At many sites in Scotland and Wales for instance wind and water are relatively reliable and plentiful resources Similarly in areas like the south western US sunshine is practically guaranteed most of the year Also the weather related behaviour of renewable sources often coincides with weather related demand For example in places like California with sub tropical climates the output from solar PV matches the demand from air conditioning nicely Similarly in countries such as Scotland wind power increases in winter in line with the increased heating and lighting demand What is more modern weather forecasting is often reliable enough that the output of wind solar hydro generating schemes can be predicted However the vagaries of the weather mean that the problem can never be eliminated entirely and it is fairly certain that networks including large amounts of renewable generation will need to cope with more extreme and unpredictable variations in supply either by using more reserve capacity or by having some means of storing energy when there is a surplus and releasing it when there is a shortfall 13 Management of storage systems like these would also fall under the remit of dispatching 2 7 The future of dispatching The main problem with current dispatching schemes as described previously is that present algorithms are based on taking information about every generator in the network back to a single
235. rent values into the timer This is what the frequency locking routine does it tries to adjust the timer value so that 128 timer interrupts happen in the space of one zero crossing interrupt In order to get finer frequency 195 resolution the timer value is dithered by adding one to it on the first interrupts in every cycle of 128 0 lt lt 128 The frequency measuring works by keeping an eye on n and the timer value which assuming the frequency locking is working form a 13 bit number inversely proportional to the mains frequency This is compared to high and low limits and a violation counter is incremented every time the limits are broken This counter 1 slowly decremented all the time so that in normal use it will not reach the threshold But if the rate of violations becomes excessive the threshold will be broken This causes the inverter to shut down immediately Shortage of I O pins mandated some dirty tricks here The same pin that drives the input multiplexer is used for shutdown the multiplexer only needs narrow pulses to operate it and the shutdown circuit is deliberately made too slow to respond to these Undervoltage on the DC bus which is potentially catastrophic also causes a shutdown The undervoltage detector circuit connects to a pin that is normally the LSB of the D A converter A series resistor is used so that in normal operation the PIC pin overdrives the detector output and the D A works properly
236. ries across the DC bus as are S3 S4 If both switches in either pair should turn on simultaneously the result is a short across the DC bus Even if this only happens for a matter of microseconds a very high current can momentarily flow which leads to inefficient operation and possible damage to the circuit To avoid this destructive shoot through it is normal practice to arrange a small delay in the turning on of each switch so that the previous one has ample time to turn off The precise nature of the delay required depends on the switching speed of the power circuit which will be discussed in more detail in a subsequent section For now suffice it to say that by the nature of their driver circuits S1 and S3 will 171 naturally turn on and off somewhat slower than S2 S4 Therefore it is sufficient to delay the turning on of S2 S4 The logic expressions for these can be rewritten S2 Q G Eq S4 Q G Eq A 8 Where Q is the same as P but with a delay introduced in each low to high transition Components R18 and introduce this delay in the circuit Both switching logic and PWM generator are shown in Fig A 9 The grid voltage signal G is derived from U2B and associated components These act as a differential comparator measuring the voltage between live and neutral of the mains input It can be thought of as a differential amplifier with very high gain so that the output is a square wave C4 C5 R14 R16 at
237. rinciple as a synchronous circuit in digital electronics and the call for bids is analogous to the clock signal in such a circuit 4 5 Proving the concept Now that some theories have been advanced of how automatic distributed dispatching might be made to work they should be tested for validity The first step would be to state clearly what the goal of the project is in terms of a hypothesis which may be proved true or false In this case the hypothesis might be The REDMan system outlined in this chapter will perform automatic generation control of large numbers of distributed electrical generators such that resources are utilised efficiently and users of the electrical power receive acceptable quality of service The next step is to design an experiment to test this hypothesis and it is here that matters start to become more complicated The obvious experiment would be to assemble an electrical grid with a large number of embedded generators and test the system on it paying attention to criteria such as efficiency voltage and frequency fluctuations outages etc Of course this is not very practical So perhaps an alternative would be to model such a network using power system simulation software or a dedicated power systems simulation computer 49 However a full size network might contain thousands or millions of generators and similar numbers of embedded dispatchers etc The computational effort required for solutio
238. rk is concerned as much with these latter issues as with dispatching itself and so it seems appropriate to continue with investigating them Due to the modular nature of the system it should be possible to drop in a more advanced economic dispatching algorithm at a later date For now though the dispatcher is serviceable and can be used in testing of the communications layer 5 6 Communications Now that the dispatcher is operational the next step is to get the information flowing between it and the sources and demands As mentioned earlier the obvious way to do this is over a computer network and for the sake of availability and ease of use there is a strong bias towards using PC compatible computers an Ethernet retwork the TCP IP protocol and the communications libraries of LabVIEW in this experimental context 5 6 1 Ethernet networking The Ethernet network was chosen because of accessibility It is the most popular standard in locakarea networking Almost all modern personal computers can be fitted with a low cost Ethernet network interface card NIC and all of the personal computer operating systems in use today have support for Ethernet hardware Ethernet supports a great variety of communications protocols including the popular 63 TCP IP as used by the Internet and of more relevance LabVIEW s communication VIs In other words LabVIEW programs can easily communicate with each other over an Ethernet network Co
239. rogramming and seasonal heating lighting requirements This system is satisfactory at present but it should be borne in mind that the patterns are predictable because they are the mean of the activities of a large number of consumers It is possible that as the sample group is made smaller as in a hypothetical scenario where dispatching is distributed its behaviour may become more random and prediction accordingly more difficult 12 2 6 2 Coping with the unexpected Prediction is never perfect so there will always be some differences between predicted and actual demand Because of the size of a national grid system these tend to develop relatively slowly and are not usually a problem For example if demand begins to exceed supply the angular momentum of the rotating machines will make up the extra power Therefore the generator speed and hence frequency of the whole network will begin to decrease As this happens automatic control systems will ramp up the mechanical power input to restore the frequency Different types of power plant vary in how quickly they can respond So operating within these constraints the dispatcher has to make sure that there is always enough plant online with enough spare capacity and the capability to modulate its output quickly enough to deal with these unforeseen demand changes Of course this assumes that the system is working properly If there is a sudden violent disturbance such as a
240. rom another program on a remote machine via DSTP This last function is of particular interest because it is exactly what is required a way to transfer numbers between programs which may be on different machines DSTP works on a client server model The DSTP enabled program is always the client To send a data item it communicates with a separate DataSocket server program and places the data item onto the server A program wishing to receive the item connects to the server and picks it up When an item is read from the server it stays unchanged on the server in other words once you have written an item you 64 may read it several times before writing it again and each time it will return the same value A good analogy is to think of the DS server as a noticeboard where a write operation pins a notice onto it and a read operation inspects the notice without removing it from the billboard Data items are identified in two ways 1 A Uniform Resource Locator URL which works in just the same way as the HTTP URLs aka web addresses familiar to any Internet surfer except that it begins in dstp instead of http 2 An attribute which is a text string associated with an individual data item This is used in situations where it is convenient to store several data items at one URL Attributes are optional and need not be used DSTP DataSocket shows some promise for this application The only foreseeable problem was that it seem
241. rst step was to draw up a loss budget A 11 1 Loss budget Up to 10 of the power can be lost Now due to the design of the inverter s output circuit every component passes the same current This is assumed to be 50 A 600W at 12V the reason for using 12V will be discussed later So the maximum allowable circuit resistance is the value which will dissipate 1096 of 600W when passing 50A From Ohm s law this is 0 024 181 Now it is assumed that half of this resistance is in the transformer primary and secondary lumped together So all the other parts must come in at under 0 012 A 11 2 Transistors UltraFets made by Intersil were prime candidates They are inexpensive devices with a very low on state resistance only 0 006 Four were used in parallel in each switching position a total of 24 devices Since the current flows through two sets of switches in series the total resistance will be 0 003 A 11 3 Current sense resistor A commonly available 30A 75mV meter shunt was chosen This has a resistance of 0 0025 A 11 4 Wiring The power connections were specified as copper sheet 2mm thick by 12mm wide The total length is about 150mm The resistance of this is 0 0001 small enough to ignore altogether The inductance was not but that is another story A 11 5 Transformer Bearing in mind the requirement for a clean magnetising current a custom transformer had to be constructed An off the
242. rticipating loads and the rest of the grid Hence it is impossible to measure the power flow in it directly and so the other two flows must be measured instead inverter power output and demand It would be possible to separate participating loads from the grid so that the import export could be measured directly but there are certain advantages to not doing so In effect it creates a virtual power network superimposed on the real one an idea reminiscent of the privatised electricity market where it is possible to buy electricity from half a dozen different suppliers but whichever you choose it nevertheless arrives at your house through the same wiring The power is only separate from an accounting viewpoint Of course there are also disadvantages the most important being that each participating load or group of loads where possible must have its power demand sensed and reported individually This requires each load to be connected to the 78 computer network In this age where more and more appliances are being enabled with on board microprocessors and networking technology is finding its way into even domestic premises this is not as serious as it may seem In fact there is interest in connecting electrical appliances to computer networks for reasons which have nothing to do with energy systems home automation web enabled appliances etc However this in turn requires that the load should have knowledge of its o
243. s one corresponding to each control or indicator of the original front panel When a VI is used as a subVI its icon appears on the block diagram of the parent VI and by connecting wires to the terminals data is passed to and from the sub VI G 9 Documenting LabVIEW programs As a consequence of the graphical model LV programs or VI s are rather more difficult to reproduce in print than conventional listings There are two main 244 problems First is that it is sometimes impossible to display all of the program in a single picture In some cases several elements must overlap and so only one of them can be seen at a time LabVIEW s documentation generator works around this by printing all the hidden parts separately but this can still be confusing since it is not always obvious where these parts were hidden in the first place Also where subVIs are used very often in practice each subVI appears only as an icon The block diagram of the subVI must be listed separately which again may be counter intuitive The second problem is that colour is sometimes important to the understanding of a program LV datatypes are identified by colour for instance strings are pink Boolean true false values are green and floating point real numbers are orange If this had been a colour publication you could have seen that all the objects in Fig G 2 were pink Colour is not indispensable however for instance the manual that comes
244. s of URL generation opening connections transferring data and closing connections Synchronisation was implemented using the previously discussed system of flags but was later modified to allow different programs to run at different timesteps Variable time stepping such as this can be a very useful tool in maximising computational efficiency but it requires careful and informed choice of the timesteps since the choices made will have implications for dynamic behaviour and stability These issues will be explored in a future section 5 7 5 Latency quality of service etc Of course the previous discussion assumes that the data passes through the network without corruption and arrives on time at its destination As can be appreciated this is often not the case Ordinary TCP IP does not have guaranteed latency or quality of service so there is always the possibility of data arriving late or never If this happens in the experimental system the value from the previous timestep will be substituted for the missing one This is a very rudimentary error correction scheme but proved to be adequate since the networks used in experiments were reliable enough and had low and repeatable latency see 5 6 3 However this could not necessarily be depended on and practical systems would require better means of dealing with errors and excessive variable latency Some newer networking protocols like ATM and Ipv6 allow creation of streams with
245. s Britain hundreds of generating stations mostly sized between 100MW and 1GW feed three phase AC power into a national grid of high voltage 275 400kV transmission lines which interconnect generators switching stations and demand centres The grid system provides redundancy so that if lines or generators fail power can be quickly re routed from elsewhere The sheer scale of this system is impressive for example the UK s national grid handles approximately 327TWh of energy per year through more than 7 000 km of transmission lines 6 This may seem very impressive but just because something is successful does not mean that it is perfect and it is now acknowledged that centralised generation has one major drawback It makes CHP almost impossible This is because electricity can be transmitted easily over hundreds of miles but there is no comparable system for transmitting low grade heat from the place where it is generated to the points of demand This is not its only failing though centralised generation can also have an effect on reliability If generation is concentrated in a few large units then the failure of one of these units will have a greater impact on the network To protect against this network planners use spinning reserve A power plant or plants of total capacity equal to the largest single plant in use is kept idling and ready to pick up the load immediately This protects against the failure of any si
246. s in the means of reckoning the power being consumed and transferring the dump commands to the load The latter is a particularly touchy subject It seems reasonable to assume that consumers like to have control of their appliances and resistance can be expected to anything that turns appliances on and off by itself It may be less objectionable in the case of equipment like dump loads inverters washing machines and heaters but appliances like computers and TVs should probably be considered sacred and no physical means of dumping them provided In these cases load dumping will probably have to be reduced to a token gesture by displaying some kind of warning to the consumer or paying such a high price that there is no chance of dumping or by incorporating a haggling procedure in the agent so that it meets dump commands by upping the price until the commands stop 7 5 1 The original The first demand agent created had no link to a physical demand at all It simply accepted power and price values typed into it This was adequate for development purposes when testing the dispatching algorithm At this stage when the whole system was just a virtual curiosity there was no requirement for the flows of actual electric power to match the flows seen by the dispatcher nor indeed to actually have any hardware at all This simple agent remained useful for loads that were constant in power draw and were turned on and off very infrequently It was u
247. s of the network The results post 227 processing was altered to disguise this fact and make it look as if the power flowing through the inverter stayed within the network instead of leaving it and entering again E 4 Adding new power components The inverter itself was modelled as a lookup table of efficiency vs power 6 points were specified representing 0 20 40 60 80 and 100 power A linear interpolation algorithm was used to find the efficiency at whatever the power level happened to be E 5 Results The inverter model was tested on an existing ESP r exemplar which included PV the ELSA Building The efficiency vs power lookup table was taken from experimental tests of a Mastervolt Sunmaster 130 grid intertied inverter Since the PV capacity of the exemplar building was very large approx 15 kWp the Sunmaster model was simply scaled to 15kW This is not strictly speaking very accurate since a larger inverter would probably have a slightly different characteristic however it was adequate for the purposes of testing the simulator The inverters were simply added in series with each PV array Fig E 1 shows an example output from the simulation the inverter power output top trace is shown along with the losses bottom trace over period of one day The small blips at around 06 30 and 17 30 were unexpected They were most probably caused by a bug of some sort Unfortunately further development and debugging
248. sampling the instantaneous voltage and current waveforms and multiplying each pair of voltage and current samples together to give an instantaneous power The time average of the instantaneous power gives the real power Both of these algorithms are normally run on a dedicated instrument sometimes including a microcontroller and A D converter or alternatively analogue multipliers or digital rate type multipliers A power meter built in this way normally costs a few hundred pounds Power meter ICs are also available that integrate most of the above functions and can be used with a timer counter or microcontroller However it was brought to my attention that most personal computers already have the proper hardware to make a power meter The sound system on a typical PC Macintosh etc has two 16 bit A D converters for stereo sound input which could be used to sample voltage and current instead The CPU on modern computers has ample power to perform all the required calculations It seemed that a PC could be used as a power meter with just voltage and current transformers to provide isolation and reduce the voltage to a safe level 217 0 1 Design considerations 16 bit conversion gives a theoretical accuracy of one part in 64 000 which is far better than required for power measurement Unfortunately though the rest of the sound circuitry in computers often does not match up to this level of accuracy It is common practice to use compone
249. se of its greatly increased complexity the proposed control centre might be more likely to break down than present day systems Unfortunately as may be seen in the following section there do not seem to be any more workable alternatives to this system at present 2 7 1 Alternatives to classic dispatching This section is based on a review of various EG systems currently available on the market The manufacturers published data for example 7 8 suggest that when designing their systems they make one of two simplifying assumptions 1 The generator is designed to be the sole source of power It is governed so that it supplies the correct amount of power to keep the terminal voltage and frequency within limits These standalone generators are simply not designed for connection to a grid their output appears as a voltage source of low impedance making power flow into the network unstable and very difficult to regulate 2 The generator is designed for network connection but the decision of how much power to generate at a given moment is left up to the end user thus sidestepping the dispatching problem completely These dumb generators will typically not work unless they are connected to a large stable electrical network NB Some generators e g 8 can operate under either of the above modes the selection being made by the user A large scale deployment of standalone generators would be inadvisable to say
250. sed to represent the consumption of the datalogging computer which had been measured as a constant 50 watts 117 7 5 2 Smart Socket The next type of agent provided a rudimentary link to an actual appliance It was still assumed that the appliance would be of the either on or off kind in other words that whenever it consumed power it was always the same amount of power The hardware used with this was a modified 13 amp extension lead It was equipped with a relay so that the power could be turned on or off by a digital logic signal and also had a sensor that could determine if the appliance was drawing current although it could not measure how much current and output a logic 0 or 1 accordingly There was one complication in order for the strategy to succeed the dispatcher would need to know if the appliance was turned on even if it had been dumped To do this a small amount of current was allowed to flow even when the relay was open by means of a capacitor across the relay contacts The sensor was made as sensitive as possible The hardware is described fully in appendix C The inputs and outputs from the Smart Socket strip containing 4 sockets were brought into the computer using a 24 line digital I O card The agent program was essentially the same as the basic model described above except that it reported the power consumption to the dispatcher as zero or the typed in value according to the state of the current sensor a
251. sequences Aggressive antrislanding protection like this can never be a part of any plan for significant penetration of embedded generation So what protection scheme was opted for Considering the situation as discussed here and in previous chapters it seemed that the only pressing need was for a rudimentary loss of mains detection The microcontroller chosen limited the options rather because there were no I O pins left and no onboard analogue digital converter so fitting AC over undervoltage trip would have been a very fiddly business Eventually the method settled on was a sensitive over underfrequency trip this was a firmware job and needed no hardware mods which resulted in instant loss of mains tripping in every scenario tested A 16 Notes on firmware The firmware has three major jobs to do locking to the mains frequency scaling the output according to the power command and measuring the frequency for protection purposes The frequency measuring and locking is the most complicated part It is based around an interrupt which is triggered at every zero crossing of the mains voltage The interrupt pin is normally edge sensitive and a programming trick is used to make it sensitive to both positive and negative edges Another interrupt is triggered by the onboard timer and causes a D A conversion to happen The frequency of this interrupt is nominally every 192 instruction cycles but it can be changed by programming diffe
252. sipated more and the inverter used more The errors and corrected values still sum to zero rounding errors notwithstanding because they were generated under the assumption that 0 still holds This assumption is discussed in Section 6 3 2 137 The effect of this divorce between measurement and reality is somewhat involved Firstly the measured results were accurate 0 3 of FS by the standards of everyday instrumentation The large accumulated errors 20 shown above are an unfortunate consequence of the operating conditions the system spends a lot of time under conditions that are only a fraction of full scale i e night time cloudy conditions inverter load of only 996 and so the error is magnified accordingly Although these discrepancies are large they mostly had no observable effect The inverter is controlled as a function of its output power so this is always correct The dump load will get a little hotter but this is hardly obvious The only obvious symptom would be an inaccuracy in the battery s estimated state of charge the real battery would discharge faster than the model The effect will be more pronounced under conditions of low RE supply and demand almost as great as supply which are the exact conditions under which these results were taken 9 4 Inverter re test The inverter was re tested to determine if the efficiency had deteriorated or the dispatching calibration had drifted since the original test R
253. source voltage is low and the operating frequency is high NB point should be made here on terminology The functions of voltage changing and DC to AC conversion are normally combined in the same apparatus Even though it is made of an inverter and a transformer or even two inverters a transformer and a rectifier if a DC DC converter is used it is customary to refer to the whole apparatus as an inverter To muddle matters even further the digital logic gate performing the function is also known as an inverter even though it has nothing to do with converting DC to AC 159 i AC Out Fig A 1 Basic inverter circuit MERE AC Out Fig A 2 Inverter with transformer Il Low voltage Out DC In Inverter 1 Step up Rectifier transformer Inverter 2 Fig A 3 Inverter with DC DC converter 160 A 5 3 Controlalgorithms From the point of view of control these different inverter circuits are more or less interchangeable The precise details of exactly which switches must be operated vary from circuit to circuit but the scheme that controls wAen they are to be operated tends to be the same The most basic is the algorithm described earlier To generate one half cycle of the output the inverter is switched to one polarity for the next half cycle it is switched to the opposite polarity This generates a square output waveform whose peak amplitude NB peak and RMS are same
254. sssessssssssssssssessssessecessecsssessssesoes 240 1 WHAT IS LAB VIEW ss AW EHE EET RE SU peto eee Pee PNE CORDES 240 G 2 WHAT IS IT USED FOR iiie hee e Eee ettet te ee e ee eto rper ve i d 240 G 3 BASICS OF LAB VIEW PROGRAMMING creer 241 GA DATA TYPES t nri eee 243 G 5 CONDITIONAL EXECUTION 243 G 6 MATH FUNCTIONS a etr RARE eR 243 7 MULTITASKING 244 8 MODULARITY ice ter EAE RERO UI Baka ee ARE RENE 244 G 9 DOCUMENTING LAB VIEW PROGRAMS eere nennen 244 G 10 REFERENGES 245 APPENDIX LABVIEW PROGRAMS 246 1 THE TESTBED 246 2 BLOCK DIAGRAMS SOURCE OF SELECTED VIS 248 H 2 1 BatMan battery management bm02 vi eese entente tenente nennen 248 H 2 2 BatMan state of charge estimator bmestimator vi eese 253 H 2 3 Smart inverter management 7 255 H 2 4 REDMan dispatching server 5 71 261 H 2 5 REDMan dispatching algorithm matching 15 vi essere 266 xiv Preface This Ph D started out with only a
255. st keep its per unit price fixed and decide for itself whether it is economic to run or not Again this may be a case for embedded prediction or it may prove unsatisfactory Third there is no prediction No attempts are made to foresee changes in demand as current AGC systems do The reasoning behind this is that more and faster control movement makes prediction unnecessary There is also no load scheduling capability i e a way of reserving a block of energy in advance Many demands represent a commitment that must be carried through and this simple system has no way of knowing in advance what the total oost of such an operation will be so it cannot help to decide whether to start the operation Load control would probably be limited to load shedding in cases of excessive uneconomic demand Fourth it is assumed that sellers are capable of providing the amount of electricity ordered by the dispatcher at short notice This presupposes that the generator s 41 response to changes in setpoint will be fast which was substantiated earlier in 3 7 5 Fifth and linked to the previous requirement There can be no consideration of requirements for minimising control movement Generators will be compelled to respond to orders for power which may change at any moment This issue is often encountered when working with large generating plant where rapid and frequent changes in power level cause thermal stress which can lead to early failure
256. st timestep F ou set and inverter efficiency Revised algorithm E E OUTPUT fw T Ka Server URL lod oad no ellefficiency estimate cE a c Pv Server URL gen side domain xi mS OPEN DC INPUT Lost stop dc input last timestep ED BuyPrice lnverterInputPower 256 Hidden parts of block diagram E 0 Default P n this mode you specify both prices and the inverter shuts down if operation is une Control actual DC input to be same as available power RP 0 0 a 257 i Economic dispatching computes buying price based on selling price set and inverter efficiency t n a A n o lt rm gen open vi C MMy Documents Vlabview backup labview labview phd realthing rm_gen_open vi oad open vi y Documents labview backup labview labview phd realthing rm_load_open vi a 21 iv 21 Q en_update vi Documents labview backup labview labview phd realthing rm_gen_update vi m PS ES d z K 3 oad update vi y Documents MV labview backup labview labview phd realthing rm_load_update vi tp 3 e c a nv gen close vi y Documents labview backupMabviewMlabviewNphdNrealthingNrm gen close vi gt e 3 uv A 258 el lt x rm load close vi
257. stem However for initial experiments the procedure described above proved to be satisfactory 5 7 3 Timing and synchronisation So far there has been much mention of timesteps without any substantial discussion of how these will be defined and enforced Fundamentally it is obvious that the timestep must be long enough to allow all of the data to be transferred and all of the calculations to be completed It is also desirable that the timestep should not vary since it would complicate conversion to and from REDMan s native energy units of watt timesteps The several different phases of data transfer will also need to be co ordinated so that one program will be sending when the other is expecting to receive These issues can all be resolved by an appropriate synchronisation system which is generated by the central dispatcher There are three phases of data transfer 1 The agent programs all submit their data to the dispatcher 2 The dispatching algorithm operates on the received data 3 The results are returned to the agent programs These phases are synchronised by a call for bids signal sent from the dispatcher to all agent programs 5 7 4 Realisation of the system Constructing the communications system according to these principles was merely a case of coding and the details will not be discussed here The end product was a 69 toolkit of subroutines or subVIs in the parlance of LabVIEW to perform the task
258. system Ross J N et al Solar Energy vol 69 no 3 pp 181 190 2000 2 mathematical model for lead acid batteries Salameh Z et al IEEE Trans Energy Conversion Vol 7 No 1 pp 93 97 3 Design and build your own electric vehicles Hackleman M Earthmind Press 1974 4 The Sinclair C5 Electric Vehicle Milner PJ et al Online at http www sinclairc5 com technical paper c5paper htm 5 Instructions for TrrMetric Battery Meter Bogart Engineering Inc 1997 Online at http www bogartengineering com TM2Instrptl pdf 6 Owner s Manual Part 890015 Cruising Equipment Corporation Seattle USA 1998 Online at http www xantrex com DocumentDepot manuals link10 pdf 122 10 Lead acid batteries Bode H E Wiley 1977 p 334 Digital Fundamentals Floyd T L Macmillan New York 1994 Feedback Control of Dynamic Systems Franklin et al Addison Wesley 1994 pp 182 183 Manufacturers websites e g http www socomec fr http www brandelectronics com 2002 123 124 Chapter 8 Experiments This section describes the experimental work the actual work that was done the reasons for doing it and the results Due to the nature of this project though it was difficult to separate these three categories Much of the experimental work was really development in other words the kind of experiments that test whether an item of hardware or software is fun
259. system 15 studied separately from all others See 2 for examples However a little consideration of the matter may give reason to believe that the various control systems interact These interactions are intentionally minimised in current power systems but may become more significant in EG systems How this will happen is only beginning to be studied see e g 3 3 2 Load frequency control This is the most basic form of control in power systems The idea behind it was hinted at in the previous section and can be restated in more detail here Supply of electricity must always equal demand In an alternating current system the frequency of the AC is an indication of how good the supply demand match really is If demand is greater than supply the generators will begin to slow down and the frequency will fall etc LFC senses any deviation from the nominal value of frequency and changes the mechanical power input to the generators thus changing the real power injection to the network in order to restore the frequency to what it should be 3 2 1 Simple governor The simplest form of frequency control where rotating generators are involved is to equip the engine turbine prime mover with a governor that adjusts the fuel supply to hold the rotating speed constant This is the norm for many smaller types of generator However in modern large National Grid type networks the scheme 1s rather more complex 19 3 2 2 Implementation
260. t from an engineering perspective it is obvious that a lot of energy is being wasted in electricity generation and transport Unfortunately this is for a fundamental reason engines which convert heat to mechanical work whether they be petrol engines in cars or steam turbines in electric power stations have a theoretical maximum efficiency imposed by the Second Law of thermodynamics The theory behind this is well known and will not be repeated in detail here Suffice it to say that in a vapour cycle heat engine the theoretical limit is the Carnot efficiency 72120 1 2 1 In modern thermal power stations the materials used will stand source temperatures of around 800 K and by rejecting heat straight to ambient a sink temperature of around 300 is possible This equates to a Carnot efficiency of 63 Of course as any text on thermodynamics will tell the Carnot efficiency can only be attained if all heat transfers in the cycle are reversible which means that they must take place across zero temperature difference and therefore be infinitely slow and hence useless In practice irreversibility caused by heat transfer across finite temperature differences mechanical friction etc reduces the conversion efficiency to 35 40 For example 5 quotes a mean efficiency of 37 for power stations in the UK With regard to transport interna combustion engines are somewhat different their theoretical benchmark is
261. t step was to design a practical electronic circuit around them The main problem was to find a way of sensing inverter output current for the bang bang control The nature of the challenge was this In accordance with feedback control theory the error performance in a system of this kind is mainly limited by the sensor The sensor would have to be accurate to within a few per cent to meet the specification with a low DC offset being vital to prevent DC injection to the line 163 MODEL bang3 DATA db INPUT IL VDEM VGRID OUTPUT 51 52 53 54 dc supply 51 52 53 54 HISTORY S1 52 53 54 IL EXEC dflt 21 dflt dflt 0 1 0 0 if VGRID gt 0 then if then L gt VDEM db 53 0 54 1 endif if IL VDEM db then 53 1 54 0 endif ENDEXEC ENDMOD EL MOSFET switch filter subber Fig A 4 ATPDraw schematic of the inverter model dead band magnitude amps inductor current demand current to TACS switches first do left hand bridge 50 Hz now do right hand half hysteresis Fig A 5 ATP MODELS description of bang bang controller 164 te 400 300 li 200 Voltage V e 300 Fig Simulated switching waveform close up shown note
262. t translation between the different timesteps of adjacent domains rather than simply passing the data across as it did in experiments 10 5 5 Concurrency and commitment handling Really the issues involved in multi domain operation are just different aspects of the concurrency problem Concurrency was dealt with across single domains by using synchronous operation However it is probably not realistic to aim for synchronous operation across an entire mult domain system So what is needed is a different way of handling concurrency The hierarchy system discussed in 10 5 4 above may be a suitable solution Another related issue is that the commitment period associated with an agreement to buy sell power may change between domains since it is always one timestep and the timestep may be different The responsibility for handling this falls onto the inter domain connection program It may be called upon to meet commitments to supply for say 5 minutes when the domain it is buying from only has a commitment of 10 seconds In order to always guarantee meeting the longer commitment period therefore it must have access to a store of energy either directly controllable by itself or for sale elsewhere in the system Alternatively the possibility must be allowed for that commitments could be broken 10 5 6 Penalties If this possibility is entertained then there would have to be a formal system of penalties for failure and a means of c
263. tated more formally In a practical system there will be many sources of energy each with its own cost the money and many demands for energy each with its own value The task then becomes one of maximising the ratio of total value to total cost Now this ratio is easy to calculate but it is somewhat harder to find a simple way of locating the maximum value There are four main computer based approaches to problem solving and decision making The first applicable where the problem can be stated as a system of mathematical equations is to use a numerical solver There are formal numerical methods for optimising any system of equations but they are sometimes computationally intensive and are not always very robust The remaining methods are applicable to decision making tasks normally done by humans which are more abstract hence not posable as a system of equations These fall within the realm of expert systems or artificial intelligence 39 The question is which one is best for dealing with millions of tiny power sources If the earlier arguments in this work on redundancy and emergent properties are to be believed the answer would be the one which is simplest and can operate with the least information especially where this is information concerning the state of physically remote parts of the network In other words the simpler the algorithm is and the less it knows about its world the less likely it is that its operat
264. te the relay D2 is a flywheel diode which prevents the relay coil generating turn off spikes 5 volt power for the logic chips and relays is supplied by the I O card 2 4 LIVE IN U2A 1 6 74HC14 U1A R2 1 6 74HC14 100k D3 D6 20 Amp 60 Volt Bridge CONTROL_OUT D7 IR 20ETSO8 1n414 1 cl 0 47u c2 0 1u 250V CONTROL IN 10K LIVE TO LOAD Fig C 2 The smart socket unit Fig C 3 Internal view of smart socket unit 215 Fig C 4 Circuit board containing opto isolators one shots and relay drivers Fig C 5 Switching sensing circuit showing relay diodes and capacitor 216 Appendix D Power measurement by soundcard As part of the program of experiments there was often a need to measure the power drawn from the mains by electronic equipment with reasonable accuracy in a convenient way that did not require any expensive hardware For accuracy s sake the power measurement had to be what is sometimes called true RMS power in other words it had to take account of power factor and the effect of non sinusoidal line current and voltage waveforms There are two ways of getting this true RMS power measurement The first involves computing the RMS voltage and current and the power factor angle found by measuring time interval between voltage and current zero crossings The product Vrms Irms cos then gives the real power The second involves
265. tem The specification has already been quoted and is quite clear on all points except for loss of mains protection This seems to be an additional category apart from voltage and frequency limits Since the G77 spec was written this has been defined more clearly in the American standard IEEE P929 2 It is now formally known as anti islanding protection Islanding is the condition where a network containing embedded generation and demand comes disconnected from the rest of the grid because of a blown fuse or tripped breaker and continues to function on its own with the embedded generation supplying the local demand The implications of this have been discussed in previous chapters it is assumed for the present discussion as the electricity authorities do that it is an undesirable condition and must be prevented The first line of defence is to have over undervoltage and frequency trips which work because inverters and ours will be no exception look to the grid to determine the voltage and frequency If the grid is cut off the inverter will lose its timing and the frequency will go wrong And if the load does not match the inverter output the voltage will go wrong too Either case will result in a trip Unfortunately there are conceivable situations where the load does indeed match the inverter output The REDMan system is one example Worse there are combinations of inductive and capacitive loads which will resonate at the
266. tenuate the mains voltage to a safe level A 8 3 Drive circuits There are a few peculiarities involved in actually applying these drive signals to MOSFETs which will be explained here A certain amount of specialised drive circuitry is needed and the circuit used is shown in Fig A 11 to which the following explanation refers The first difficulty is that the gate of a MOSFET has considerable capacitance It is essential to turn the MOSFET on and off as quickly as possible and so high peak currents are required to charge and discharge the gate capacitance In order to provide the current a complementary pair of transistors Q1 and Q2 are used as emitter follower buffers These are special transistors designed for the application and can supply peaks of up to 2 A MOSFETs M5 and M6 do not have this drive circuit because they only operate 50 times per second and so do not require high speed switching The second difficulty is that the drive signal must be referred to the MOSFET source In the case of M2 M4 M5 M6 this is not a problem since the source is grounded But M1 and MG have their sources connected to the output terminals When they are turned on the source voltage will rise to the DC bus voltage and so to keep them switched on the gate voltage must be kept higher than the DC bus This is taken care of by a bootstrapping circuit which supplies the gate drive voltage from a 172 Uy 9402 3 21412 25g uo s
267. the least An EG system made entirely of these each feeding its own demand centre would throw away all the advantages of a grid It would be very inflexible if a 15 demand centre required more power than its generator could supply the only solution would be to buy a bigger one Utilisation would be very poor too the generators would have to be sized to supply the peak demand and in most cases the mean demand is much less Connecting large amounts of dumb EG to the existing network would be equally unwise because it would lack any form of dispatching There would be no guarantee that supply would equal demand 2 8 Conclusions By reviewing the way in which electricity generation works at present it was seen that relatively large improvements in efficiency could be made by aggressive use of CHP But to make CHP work embedded generation is essential There is reason to believe that the full benefits of EG will not be realised unless it is network connected and suitably dispatched to ensure stability and good economic use Unfortunately it seems that the dispatching system presently used to control large centralised power plants will not be practical for millions of small generators nor will the controls fitted to present day EG solutions Much work has been done on the technical and economic implications of EG and the need for a new kind of dispatching system has been recognised 14 but there is no evidence of anyone having
268. tialise Default PF m ist of SubVIs bmstatemachine vi C MMy Documents labview backup labview labview phd batman bmstatemachine vi bmestimator vi C My Documents labview backup labview labview phd batman bmestimator vi bmchargemeter_sub vi C MMy Documents labview backup labview labview phd batman bmchargemeter_sub vi et l d bm time vi C MMy Documents labview backup labview labview phd batman Mom time vi bm volt vi y Documents labview backup MlabviewNlabviewNphd batman Mom volt vi rm gen update vi Documents labview backup labview labview phd realthing rm_gen_update vi a S oad update vi y Documents labview backup labview labview phd realthing rm_load_update vi tp 3 3 nv gen open vi y Documents labview backup labview labview phd realthing Nrm gen open vi m gt z 3 21 250 nom qm MES a E Pel Cel E rm gen close vi C MMy Documents V labview rm load open vi C MMy Documents V labview rm load close vi y Documents labview bmenergymeter Documen sysglobal vi C My Documen bmstatusgloba vi ts labview ts labview lvi Documen ts labview backup Ml backup M backup M backup l backup 1 backup 1 abv iew labview phd realthing rm_gen_close vi labview labview phd realthing r
269. tigated in more detail self discharge should be measured and its implications investigate and discussed Data should be gathered to check if the assumptions and control strategies built into the various agent programs are correct suitable Testing cost models This is related to the previous topic which described ways of testing if the dispatching of simulated generators worked optimally The question here is whether the simulated generator is an accurate representation of the real one There are four devices considered as generators in the experimental system the PV modules the ducted wind turbines the inverter the battery and the grid PV modules and DWTs were considered to always have zero cost The grid was considered to have a constant price representative of the price given to small consumers Neither of these assumptions is verifiable by experiment and so cannot be considered further However the batteries and inverter each had their own cost modelling built in and this can be tested 129 Inverter cost model The selling cost of the inverter output is dependent on the input price and the power loss in the inverter This relationship was discussed in more detail previously The model can be tested by measuring power loss in the inverter and checking that it is the same as the modelled power loss over the whole range of powers Battery cost model Again the selling cost of the battery output depends on the input
270. time scale 2000 1500 1000 500 0 500 1000 1500 2000 T T T T T T T T T 0 2 4 6 8 10 12 14 16 18 t ms 5 0048 XXX0030 Fig A 8 Simulated line current waveform 165 or saturation of the output transformer if one was used It would also require high bandwidth so that the bang bang control would work properly Finally it would need excellent immunity to interference from the inverter s own HF output voltage A current transformer would seem attractive because of its immunity to interference Unfortunately the sensor needs to be DC accurate and transformers do not respond to DC at all This could probably be got round by ingenious circuitry but even then the transformer would have to be physically large if it was to handle the 50 Hz component of the inverter output without saturating Its leakage inductance might then cause problems with accurate measurement of the HF components A natural contender then was the Hall effect current transformer It can respond to DC and has a high bandwidth Unfortunately though the devices investigated all had a very large DC offset which ruled them out The remaining option was the humble sense resistor This had been ruled out because it would have to be placed in series with the inverter s output In this situation the small voltage across the resistor would have to be separated from a large high
271. ting electrical grid as if it were one generator in the system because it will essentially take over control of load voltage and frequency easing the task of the other generating plant It also allows the REDMan network to be merged with the existing mains wiring so that REDMan dispatched generators and loads can simply be plugged into ordinary wall sockets which would be very convenient However electricity companies will not tolerate any equipment on their grid if they think that it might be unsafe or that it might interfere with other users This means that any experimental equipment connected to the grid should be made to comply with their recommendations Nevertheless time and budget restrictions associated with this project might make this an unreasonable objective The decision to interact with the grid or not will have a very strong influence on the testbed design and so it 1 logical to tackle this issue first 73 6 1 Grid connected working The UK electricity authorities publish recommendations which define a minimum standard and means of connection for embedded generating equipment Some of these recommendations are actually law Others are not and the electricity company has no legal way of forcing you to comply with them The legal restrictions relevant to this project see 1 can be summarised as follows It is illegal to connect your installation to the grid if it can also be supplied from an alternative source That is
272. tly and therefore it is tempting to ignore it Whether short or long run marginal costs are used the effect is exactly the same The merit order of the plants depends on the amount of power bought from each which in turn depends on the merit order etc Therefore achieving the economic optimum means solving a set of simultaneous non linear differential equations The exact form of these equations depends on the functions describing the marginal cost of each plant as a function of power 27 3 6 2 Application to EG These concepts do not translate directly to embedded generation The first problem is that EG could involve very large numbers of generators potentially millions on a national grid sized network Classical economic dispatching of all of these would require solution of an enormous equation set not to mention the logistical trouble of gathering all the data together in one place where the equations could be solved This is one of the main problems to be addressed in this thesis and will be discussed at length in chapter 4 The second problem is that the marginal cost of many EG sources is hard to define For instance in CHP plant the marginal costs of heat and electricity as well as the amounts of each generated are tied together In other words the marginal cost of heat is determined by the amount of heat you buy but also by the amount of electricity and vice versa In a situation where heat and electricity were sold in
273. to the contrary It is probably more a matter of convenience and expense than safety embedded sources of power in the distribution system would make fault location and repair much more difficult and time consuming because there will no longer be a clearly defined power flow from generator to load When the network stays live on both sides of a fault how do you trace where the fault is Of course the high voltage transmission system faces this exact challenge and its solution is to use automatic protection relays that signal when and where a fault has occurred There is a more serious side to this issue however if an embedded generator were to trip out and then restart at a later time while someone was working on the system a very dangerous situation could result The protection systems currently used with EG do not allow restarting unless a healthy voltage and frequency are already present and this is therefore one feature that should be kept The second objection is somewhat more grounded in fact but again it need not be insurmountable If the power quality within an island was of a good standard the voltage and frequency would remain very close to that of the rest of the network Therefore it might be possible to reconnect it at an instant when the relative phase was correct This phase sensitive switching would require replacement modification of many breakers and reclosers in the existing system and therefore
274. tructures in a normal programming language A single wire can carry data of any of the kinds familiar to programmers For instance it may hold an integer signed or unsigned of various bit depths a floating point real or complex number of single or double precision a Boolean true false value a text string an array of any of these of any size and number of dimensions subject to memory or even a cluster which is a user defined bundling together of any desired mixture of any of the above including other clusters The cluster is equivalent to the typedef struct in G 5 Conditional execution LabVIEW has conditional structures equivalent to all of the familiar ones in conventional languages such as do while for next case etc They appear as boxes and the code to be repeated selected is placed inside the box Case structures can be confusing because only one case is visible at a time See G 9 Fig G 4 Some conditional structures in LabVIEW G 6 Math functions Since it is used in datalogging and analysis LabVIEW has very good math functions The built in functions are the same as would be expected in any scientific language such as FORTRAN However there are many more routines or subVIs see G 8 provided for applications such as statistics signal analysis and the like User defined 243 math functions can also be written in a C like syntax and embedded in the VI using formula nodes
275. ught to be a problem with the inverter management itself and the problem was cured by a revision of the algorithm The true cause did not become apparent until the final results were examined in detail 9 7 Summary A four day long test run was performed The logged data allowed analysis of the dispatching performance terms of energy balance and transient behaviour Also the DAQ and control hardware used in the test rig was checked against other measuring equipment The results can be summarised as follows The dispatching engine itself worked very well the other software worked properly with the 145 exception of the inverter management agent There was a slight discrepancy in its operation which left a small amount of power unaccounted for causing a difference between the power flows in reality and those seen by the dispatching engine Errors in the measurement hardware also contributed to this effect However these errors and differences were not of great importance and basically the system did operate as intended In the next chapter these findings will be discussed and some conclusions drawn as to the validity of the REDMan concept as a whole 9 8 References 1 PCI 6023E 6024E 6025E User Manual National Instruments 2000 online at http www ni com 146 Chapter 10 Conclusions 10 1 Does it dispatch successfully The answer here is yes Experiments showed that the matching of supply and demand was
276. uired by the first demand is subtracted from the first source in the source list going on to subsequent sources if the first one cannot give enough power Meanwhile a running tally is kept of the per unit cost If it exceeds the cost associated with the demand then that demand cannot be accommodated This process is repeated until all demands in the list have been dispatched Then the results are presented to some sort of mechanism which will make the real generators and loads obey them 53 Get size and cost pairs P maxi for each source Sort source pairs according to magnitude of C smallest first Get size and cost pairs Cai for each demand Sort demand pairs according to magnitude of Cgi largest first Is Posi greater than or equal to Pg increment i Is P CJ less than or equal to Pay Cg refuse jth demand accept jth dem and Is j greater than nunber oj demands Send powers P to control hardware Wait for next time step Fig 5 1 Flowchart of a simple dispatching algorithm 54 This flowchart is a good enough specification to form the basis of a computer program There is just the matter of sorting numbers for this part of the program the chart indicates sort the numbers without specifying how this should be done Sorting of lists is a very common operation in computer data processing and there are many different methods in use each best s
277. uitable spot was the south edge of the roof of the University s James Weir building Next to minimise electrical losses in the wiring the rest of the DC equipment would have to be sited as close to the generators as possible And since there would be a large amount of sensitive signal wiring between the DC measuring shunts the computer and the inverter which would pick up interference unless made as short as possible it was logical to site all this apparatus in the same place That place proved to be a convenient room on the top floor of the James Weir building as close as possible to the PV module mounting location with the power cables led out through a roof window Nevertheless the cable run was still around 20 metres and so the cables had to be oversized to minimise power losses 6 4 4 Power wiring Cabling for all the DC equipment was brought to a custom made distribution board constructed from acrylic sheet with bus bars of copper strip and steel bolts for terminals The wiring diagram is shown in Fig 6 2 The generators two PV arrays and two ducted wind turbines and dump load connected to these via fuses and 10 amp current shunts The battery connected via a 30 amp shunt and an isolation switch The inverter was connected directly across the busbars with neither a fuse nor a shunt Owing to the currents of up to 30A RMS expected in these connections and the assumption for the sake of power calculation that volt drops in the
278. uited to different kinds of list One of the fastest overall is the well known Quicksort algorithm However in the proposed experimental system the lists are liable to be very small about 10 entries and it was felt that the time taken up by sorting would be insignificant compared to other aspects of the program whatever algorithm was used 5 2 Details of programming Now that the algorithm has been specified there is just the matter of what programming language it should be written in It was decided to use National Instruments LabVIEW environment 1 for reasons of convenience which were touched on earlier and because of its good and easy to use support for data acquisition and networking which are of great use in programs for communicating across networks and speaking to experimental hardware A review of the LabVIEW system can be found in Appendix G Really though the choice is a matter of convenience almost any high level programming language currently in use whether it be Java or Visual Basic could easily have managed the task in hand and the reader may well prefer to take the algorithms developed here into his or her preferred programming environment Therefore programs written in this work will generally be presented in the form of flowcharts or pseudocode Since it would take up a great deal of space the original LabVIEW source code block diagrams has only been printed for a few of the most important progra
279. ul because the result depends on which properties you choose to compare The choice of property is necessarily subjective there is no experiment or equation to solve that will indicate which are the right comparisons to make But there are obvious properties which people experienced in this field would presumably want to see tested These are 1 Does it actually dispatch In other words does supply equal demand average In the steady state Under transient conditions 25 Is it truly economic Does it manage power flows and capital items such as batteries to deliver maximum energy per unit cost to its operator How does the performance compare to existing economic dispatch algorithms How does the cost of the control hardware and the power it consumes weigh up against the benefits 3 Is it novel What does it do that was not or could not be done before 4 Does it really deliver on the promise to be usable with large numbers of generators These are the questions that the experiments must try to answer as well as possible 1 is relatively easy results from the datalog files can be scanned to check for consistency and power measuring equipment can be used to check that the hardware is transporting the same power as the software commands 127 2 is more difficult because it depends on other questions Who is the owner of the system How are income and expenditure defined Are the simplifying assumptions at the
280. ules Rule 1 Buy the cheapest electricity first Rule 2 Sell it to the consumers who are prepared to pay most for it 40 3 Continue this process until either all the consumers are satisfied or the supply runs out or the total amount of money paid to sellers of electricity approaches the total amount earned by selling it on allowing for a profit margin where one is required 4 3 2 Assumptions underlying the rules Now that the rules are established it is time to debate the assumptions under which they were created First of all and most drastic compared to the traditional paradigm it was assumed that all the generators and consumers actually are dispatchable This is totally at odds with the status quo in fact as far as some power engineers are concerned distributed generation is synonymous with norm dispatchable A large proportion of this thesis is devoted to possible ways of making distributed generators dispatchable and making renewable energy generating technologies that are not dispatchable behave as if they were Second it was assumed that a single per unit price is enough to describe the true cost of electricity This has the effect of banishing marginal cost the implications of which were hinted at in section 3 6 2 In the present context the impact of this could be minimised by making sure that the system maximises utilisation and the remainder of the responsibility shifted to the generator so that it mu
281. um OZ wona 662 102 S3 jeuuojsuea ad i 3 v AODL ACI AOPC XL ul 404851534 eAgonpuruou 15040 XZ PIaSIIW 8b LENL sqq ADOL AYOS 1324 zoordn ig EIH apis pueu ja se aues L N34 LA 1935014 euueu2 N 2 001 FOP POBENZ 58 47 5 7 93701407434 90 yo co 2 xL i 248 0 0 M m Te D olm a ta ob 32 ozz ASL SI LN deu tage One half only shown iver s f dr ircuit diagram o Fig 11 173 capacitor C3 connected to the source This capacitor is charged from the auxiliary 15 V supply via R6 D1 during periods when M2 is on and the source is grounded side effect of this circuit is that it slows down the switching action compared to the non bootstrapped version This is because the voltage at Q5 collector must swing by a larger amount and hence the Miller effect in Q5 will be greater The network C5 R12 injects extra base current to compensate for this as much as possible but it is still somewhat slower It should be noted that these drive circuits all invert the signal a HIGH input signal turns the associated MOSFET OFF This is taken care of by modifications to the switching logic A 8 4 Reference generator If this circuit is to produce a sinusoidal output it requires a sinusoidal modulating wave as a reference In order to
282. unfortunately not the case at the moment in Britain a mere 6 5 of UK electricity comes from CHP plants The reason for this neglect is that there is a serious problem with large scale use of CHP The problem is intimately connected with the way in which electricity is generated and used 2 2 1 Generation and transmission of electricity Currently electricity is generated in large power plants situated some distance from the point of demand This system of centralised generation dates from the late 19 century electrical equipment of that era used direct current of low voltage which could not be transmitted more than a few miles without incurring excessive losses in the conductors or using very thick and costly conductors The solution was to build a small coakfired generating plant on every city block which was inconvenient inefficient and expensive Soon better ways of transmitting power were invented to help in the exploitation of hydro electric resources This economical long distance power transmission brought in a massive expansion of electricity supplies and The three phase high voltage AC transmission system was first used in 1891 and still exists in essentially the same form today allowed generating stations to be made as large as technically possible in order to gain economies of scale This trend of increasing size and centralisation has persisted until the present day At present in a country such a
283. ures not exceeding 60 C ina 20 C ambient without forced air cooling Unfortunately the current shunt ran rather hotter than this since it was wedged underneath the circuit board and overloaded beyond its 185 design rating This problem was mitigated by raising the circuit board up so that air from the cooling fan could circulate underneath it The main circuit board transformer and an extra board containing mains relay EMI filter and parts of the zero crossing detector were mounted to an aluminium baseplate This was installed in a casing that had once housed a variable speed motor drive as shown in Fig A 17 The cooling fins while they look the part do not actually do anything cooling is by forced air sucked in through the fan and exiting through a grill on the opposite side of the case not visible in figure Two high breaking capacity HBC fuses were also fitted one 32 amp in the DC circuit and one 63 amp in the low voltage AC circuit The inverter was tested in this state The original electrolytic capacitors used for decoupling were found to be inadequate straight away they were overheating and there was excessive DC bus ripple enough to cause the circuit to malfunction They were augmented by six pulse rated plastic film capacitors in parallel each 1 F 63V and all the decoupling capacitors were relocated onto two copper strips right at the DC input terminals instead of being on the PCB as before This cured
284. useful So with the increasing use of rechargeable batteries in consumer products more advanced ways of determining state of charge have been developed The simplest of these is current compensated voltage measurement This relies on the fact that the terminal voltage is affected by the electrolyte concentration which in turn is proportional to the state of charge Of course the terminal voltage also depends on the current drawn but by measuring the current its contribution can be compensated for In the simplest form of this system a known current is taken from the battery and the voltage measured directly Often the current draw of the equipment itself is quite constant and only the voltage need be measured This is the principle behind the ubiquitous low battery indicator light More advanced implementations use an actual measure of the current and may also measure battery temperature since this affects the voltage too Where the current can fluctuate widely time averaging of the result is used to hide any transient errors in the calculation 4 This method can be reasonably accurate and is applicable to almost every battery chemistry not only lead acid Some battery types have very little change in voltage with state of charge though the worst 97 offenders in this respect being batteries whose voltage remains constant until the last of the charge is used up and then suddenly collapses In cases
285. ut of synchronisation and trip out If a few large generators are lost in this way the lack of generation could initiate a frequency collapse of the entire grid The stakes are high and so this is the best known and most studied form of instability 3 7 2 Classical stability studies The classical method of analysing stability described in detail in 15 involves a number of simplifying assumptions Transmission lines transformers etc are reduced to a network of shunt and series reactances Synchronous generators are represented as a voltage source in series with an inductive reactance This can be assigned one of three values called synchronous transient or subtransient depending on the expected timescale of the fault event Resistances are often neglected A steady state solution is then performed to find power flows and rotor angles of the synchronous generators in normal operation The fault is then applied Normally it is considered to be a three phase balanced fault since this simplifies calculations and is also a worst case scenario Other types of fault require a full unbalanced solution using per phase or phase sequence methods Obviously this is not a problem with modern simulation software but it still requires more effort in description of the model A fault is a short circuit and so cannot consume any real power It is assumed that the real power input from the prime mover remains constant while the fault is on Th
286. verter agent program No attempt was made to synchronise the operations of the two domains except that they were configured to have the same nominal timestep The effect investigated here is this When a change is made in one domain it may alter power flows prices associated with the inverter This will then carry the effect through into the other domain which may cause changes to take place there also These may in turn be carried back to the original domain and so on The question is how long do the domains take to converge to their new state Are there situations where they do not converge at all but become unstable What are the factors influencing this The results suggested that the behaviour of the inter domain connection was sometimes determined by the slowest component in the system In particular the battery management program updated every 10 seconds whereas the other programs updated every 1 second At times when the inverter was trading energy with the battery it would have to wait up to 10 timesteps to receive the energy that it ordered This does not comply with the specification laid down for the system which requires a response by the following timestep The effect was to make the inverter management slow to respond and sometimes even unstable as was the case with the original inverter management algorithm The inverter sometimes became trapped in a loop toggling on and off repeatedly At the time of development this was tho
287. voltage and frequency within acceptable limits for the consumer while operating the generating plant and transmission systems as efficiently and safely as possible This task can become very complicated in a large electrical network with multiple generating stations each with its own economic and dynamic behaviour connected together by a grid which has losses and constraints on possible power flows Multiple different control systems must work together without any negative interactions arising The functions performed by these control systems can be grouped under a few very broad headings Frequency control also known as load frequency control LFC Automatic voltage regulation AVR Protection Economic dispatching ED 18 LFC is the system or systems that tries to keep the frequency of supply constant It does this by increasing or decreasing the real power output from generators AVR as the name suggests tries to keep the voltage constant or at least within limits for all consumers This may be done by tap changing transformers which step the voltage up or down slightly as required or by devices which produce or consume reactive power Protection ensures that no component is overloaded to an unsafe extent by actuating circuit breakers in the event of a fault or overload Lastly ED tries to ensure that all plant is operating as efficiently as possible These systems are often analysed in isolation e g the protection
288. which in this case is free and on the expense of battery wear and tear The assumption of free input cannot be tested so it is the wear and tear model which must be examined If this can only be done by performing life tests on actual batteries it may be impossible due to limitations of time 8 5 3 Interactions The prototype REDMan system does not take full account of all economic interactions between components Some of the prices are fixed in the interest of simplifying the interactions so that system behaviour could be more easily understood and checked during development In particular the battery selling price is fixed when t could be calculated by the wear cost model The result of removing this price fixing is unknown and probably would be quite interesting so this would also be a worthwhile experiment 8 6 Examining multiple domains Some of the information required for this would be forthcoming from the logfiles What would be of interest here is the transient response in one domain to changes in the other Any sign of instability where a change in one domain caused a change in the other that then reflected to the domain where it started etc would be a cause for concern Also of interest would be the tendency of changes to ripple through multiple domains the length of time taken for this and whether these time delays are of practical consequence This is less evident in the experimental results but could be obtained
289. wn power consumption This is somewhat more difficult because historically there has been no motive for doing anything of the kind It may require extra sensing circuitry which adds complexity and cost In the case of a very simple appliance such as a lamp the extra cost may be completely unreasonable in proportion to the cost of the appliance itself It is fortunate then that for many simple appliances there is a workaround The power consumption of a resistive load like a lamp or heater is substantially constant If a 60 watt bulb is switched it is reasonable to assume that its power consumption is 60 watts In this case knowing the power consumption reduces to a simple case of determining whether the appliance is ON Another possibility would be to group simple loads in order to spread the cost of the sensing circuitry amongst them For instance taking the example of a lamp the whole lighting circuit of a domestic premises might count as one single load as far as the dispatching software is concerned and one power sensor which might conveniently be fitted in the distribution board would be sufficient Some prototype forms of demand measuring and reporting equipment were designed and built as part of this work Appendices D and E describe these in detail But to return to the issue at hand there is now a basic specification for the structure of the experimental system and it is time to tackle the issues of the actu
290. y charger ON when the battery voltage falls below 11V and then switch it OFF when the voltage exceeds 14 4V indicating that the battery is full F 5 Circuit descriptions F 5 1 Controller module Refer to Fig F 1 The input signal is first amplified by differential amplifier U1 which removes common mode interference R1 R4 set the gain of this stage while Cl C2 help to filter out noise Where a high gain is not required U1 canbe left off the board and a spare op amp used instead by fitting J3 J9 and J10 The amplified signal is fed to the meter switch via R5 and to comparators U2B U2C Here it is compared with upper and lower thresholds set by trimpots VR1 and VR2 The threshold voltages are also fed to the meter switch A stable 3 75V reference is provided by 2 5V reference diode U4 and amplifier U2D J7 and J8 select the lower extent of the trimpot adjustment to be either OV J7 or 2 5V J8 R13 R14 R15 R16 provide 0 1 hysteresis to help eliminate any remaining noise The comparator outputs are used to set or reset the S R latch U3B U3D via R17 D1 R18 D2 which convert the bipolar outputs of the comparators to an 0 5V signal U3A and associated components provide a power on reset pulse to initialise the latch to a known state Either the Q or Q output depending on which of R21 or R23 is installed is used to drive power transistor 232 Q2 predriver Q1 which also lights an indicator LED If R23 is installed then Q2 a
291. y If his plant is connected to the grid it should be possible to sell this electricity to another customer But it is also possible that as the network is made bigger it will become more complex and the scope for error will increase too As the system gets larger the dynamics get harder and harder to model It would be almost impossible to predict the transient performance stability response to faults etc of a network with hundreds of thousands of embedded generators Therefore power companies and EG manufacturers show a reluctance to move in this direction There seems to be a great lack of work on understanding the issues involved Managing an electrical grid is rather like commodity trading but fraught with difficulties because unlike other everyday commodities electrical energy is difficult to store in any significant amount Even when converted to another form of energy the possibilities for storage are limited and some energy is always lost in the conversion Therefore in practice the supply must be continuously adjusted to match the demand and it must be done quickly because the network does not have much of an energy reserve The only inherent energy storage is the rotating inertia of the generators themselves and any mismatch between supply and demand will therefore alter the speed of the generators and hence the frequency of the alternating 11 current So the first type of management required is load frequency
292. y consuming 19 watts D 5 Conclusions Measuring power with a computer soundcard certainly seems to be possible as the experimental prototype shows It requires a minimum of extra hardware and gives acceptable though not excellent with this particular soundcard results It would be useful in applications where there is access to a computer power measurements are desired but the budget does not permit a dedicated power meter 221 TX1 CT 100 1 From mains To load current_signal voltage_signal Fig D 1 Circuit of measuring unit Fig D 2 Measuring unit 222 Fig D 3 Inside unit showing current and voltage transformers CT is on left IE Power Analyzer Volts RMS Amps RMS buffer siz Fig D 4 Power analysis of 60 watt light bulb 223 ib Power Analyzer Current Volts RMS Amps RMS Volts Amps f s buffer size quit E t QUIT Fig D 5 Power analysis of mains power unit of a laptop 224 Appendix E Modelling inverters in ESP r As part of this work various computer tools for modelling renewable energy power systems were investigated One of these was the integrated simulator ESP r This was originally developed 1 for building simulation but in the course of time the capabilities were greatly extended It now aims to perform integrated simulation of all energy systems and processes at work within a building or other environmental system Of particular interest
293. y to the resistor ladder output The inverter connected directly to the computer s printer port Interface details in appendix A The one remaining connection was an Ethernet cable linking the computer to the campus network and thus eventually to another computer in charge of the AC loads 6 4 6 Testing Initial tests were restricted to checking functionality of the system components and looking for any interference between them that could not have been found when they were tested separately In this respect the biggest worry was that the inverter would generate electrical noise which would make accurate measurements impossible possible precautions were taken against having it give out high frequency switching noise but it was theoretically inevitable that in generating a sinusoidal output the current drawn from the DC bus would be a sine squared waveform Since this is only of twice mains frequency i e 100 Hz and of high amplitude 50A p p at full power filtering it down to a small amplitude would require very large filtering components and so would not be very practical Luckily a solution was at hand in the form of digital filtering By having the DAQ card take several samples and calculating the mean value a simple low pass filter was implemented in software Sampling at 1 000 sec and taking the mean of successive blocks of 100 samples was found to reduce the ripple to an acceptable level although at rather a high
294. y vector and a list of true false values indicating whether each demand is to be accepted or refused termed the dump vector These arrays are then re sorted according to the previously added index number to return them to their original order The LabVIEW block diagram of this program is included in Appendix H section 2 5 5 4 Testing Once the code had been written the next step was to test it There are two components to the testing process Most important is to show that the algorithm itself is correct in that it can perform optimal or close to optimal dispatching of a given set of sources and demands Under the assumption of zero marginal cost the ED optimisation problem is trivial Hence it is sufficient to verify that the program performs in the same way as the algorithm originally specified that is to debug it The basic method of doing this is to run the program on a sample dataset and compare the results with those given by hand calculation according to the rules from which the algorithm was derived This is a fairly unexciting process but by way of a demonstration of the program an example run is given here 56 5 4 1 Worked example The first task is to make up a list of sources Choose 3 sources and let their available powers be 9200 72 and 159 watts and their prices 0 17 1 2 and 3 units Also make up a list of demands for instance 3 demands with powers of 400 9000 and 1 watts and prices of 10 1 and

DISTRIBUTED DISPATCHING FOR EMBEDDED GENERATION

Contents

Download Pdf Manuals

Related Search

Related Contents