Home

April 2007 (pdf - Cal Poly San Luis Obispo

1. 50 45 Voc 5 AQ 150 Wp 35 high E 20 _ 2 i 90 Wp s gt ou f 20 8 Wp 15 E 46 Voatt o 20 40 60 so 100 Vv PWM Duty Cycle Figure 4 4 a PV Array Voltage Response for Figure 4 4 b Example of Corresponding Varying Insolation Levels 68W Load I V Curves Unexpectedly interfacing the current source to the DC DC converter in Simulink presented a non trivial problem The software recognizes that with the MOSFET switch closed the circuit topology presents an inductor in series with a current source This is an illegal configuration as the inductor current would not be independent It is therefore necessary to include some circuit element in parallel with the current source Various options were explored including resistors and controlled voltage sources but a capacitor 44 actually provides the optimum system behavior This is because the capacitor offers stability in maintaining the converter input voltage which is used to calculate the PV array current If the capacitor is too large the response time will be too slow and the voltage level will never rise if too small it does not provide the needed impact and the voltage will fluctuate far too much A good value for the capacitor was found by experimentation to be on the order of microfarads in this case 10uF Thus the Simulink model includes an extra 10uF capacitor on the converte
2. Figure 5 7 March 19 2007 SLO Insolation and Temperature The plot of Figure 5 8 shows the battery status measurements taken by the SuPER status system on the 19 of March The motor was run from 11 30 to 12 30 71 Vb 428 478 528 578 t current A voltage V o o a SS time min Figure 5 8 March 19 2007 Motor Operation Measurements The insolation and temperature data taken during the day are fed into the Simulink simulation The resulting estimated voltages and currents are shown in Figure 5 9 20 15 Sap pos Yue Vo En sel de me ss om l t time min Figure 5 9 March 19 2007 Motor Simulation Much of the noise that appears in the simulation plots is largely due to the coarse duty cycle resolution of 2 5 Improved resolution which would require greater simulation time would result in much more accurate levels However the general trends over time can clearly be seen These are very promising results in spite of the fact that we do not have a proper model for the Outback MX60 converter or its control algorithm 72 Figure 5 10 shows for the same time frame the battery SOC as predicted by the charge estimation code running on the laptop as compared to the simulation SOC estimate 0 98 0 96 simulation 0 94 status sys
3. Cooler 60 W LEDs sw4 45w Laptop 60 W USB 6009 sw5 Vn In Tn Motor 250 W Figure 3 2 SuPER Status and Control Interface Diagram MPPT is accomplished with the simple and commonly used perturb and observe P amp O algorithm The algorithm is presented in detail in Aki O1 s thesis 28 section 3 5 1 but briefly outlined here The purpose of the algorithm is to maintain an impedance seen by the PV array that will cause the array to output power at peak capability This is done by adjusting perturbing the DC DC converter duty cycle at periodic intervals and monitoring the resulting array power output through current and voltage measurements A negative change in the power output will cause a reverse in the direction of the perturbations a positive difference has the opposite effect Figure 3 3 shows the location of the maximum power point on the I V curve of the BP150SX solar panel at peak output 36 Module Current A Module Output Power W 30 35 40 45 o 5 10 15 20 25 Module Voltage V Figure 3 3 BP150SX I V and Power Curves 28 The status system sample period Tss currently set at two seconds puts a maximum rate on the control algorithm execution Note that this is entirely different from the NI DAQ device A D sample rate which is much higher Observation of the performance of the MX60 converter shows rapid response times under quickly changing condition
4. fit for y Lor D 4 Battery S function Code batt_voltage_src c VRLA battery model oX H int SOC initial Soc 11 out SOC2 Vbat Ed Adapted to C by Tyler Sheffield 2 14 06 LE SMITA IAAI A AAA AMAA A IAA A a define constants double SD 4 34e 5 battery self discharge rate h 1 double SOCm 1330 max battery energy Wh double ns 6 number of 2V series cells double SOC ee B temp vars adjustable parameters double k 8 battery charge discharge efficiency double t 0 0033334 time step constant equal to block sample time converted to real time resistance model adjustment multipliers double rd8 5 rc8 15 rd9 50 rc9 15 rd10 50 rc10 15 SOC coefficients double phi8 2 174 phi9 1 43 phil0 625 if fabs I1 0 gt 2 high current case requires battery capacity adjustment socm 179 68 log fabs I1 0 1435 2 taken from Excel curve mapping soc2 0 SOC1 0 B SOC2 0 line 75 if SOC1 0 lt 8 under 80 case if 11 0 lt 0 discharge mode v1 0 1 95 18 B ns R1 0 19 1037 B 14 ns rd8 SOCm else if I1 0 gt 0 charge mode v1 0 2 148 B ns R1 0 758 1309 1 06 B ns rc8 SOCm ee k V1 0 I1 0 SD SOC2 0 SOCm t phi8 soc SOC2 0 ee SOCm soc2 0 SOC ARE else if SOC1 0 gt 8 amp amp SOC1 0 lt 9 94 if 11 0
5. Adapted to C by Tyler Sheffield on 2 14 06 ANA AAA AAA AAA AAA AAA AA AAA AAA AAA AAA AAA AAA AAA AAA AAA AAA Define variables and initialize double C 0 025 Step size for duty cycle change 2 5 Calculate new Pa double Pa_new Vpv 0 Ipv 0 deltaPa adjustment offset double dpoffset 0 rag pass through values double DCnew DC 0 charge_mode_out 0 charge_mode 0 if charge_mode 0 1 amp amp Pa_new lt 3 low power state always go up DCnew DC O C Increase DC else if charge_mode 0 1 amp amp Vb 0 lt 13 7 ed bulk charge case fi P amp O Algorithm starts here double deltaPa Pa_new Ppv 0 dpoffset count 0 0 reset float mode counter if deltaPa gt 0 he keep going if DC O gt DCprev 0 DCnew DC O C Increase DC else DCnew DC O C Decrease DC else if deltaPa lt 0 go opposite if DC O gt DCprev 0 DCnew DC O C Decrease DC else DCnew DC O C SIncrease DC else DCnew DC 0 No change elseif else if charge_mode 0 1 amp amp Vb 0 gt 13 7 count 0 count 0 1 if count 0 gt 1 must read 13 7 twice to enter float mode 92 charge_mode_out 0 2 change to float mode DCnew DC O C Decrease DC a elseif else if charge_mode 0 2 amp amp Vb 0 gt 13 4 move towards disconnect DCnew DC O C
6. load of eight quarts of water It will likely be undesirable to invest the power necessary to bring eight quarts of water down to the minimum temperature We will simulate and test towards maintaining the water between 20 and 30 F below ambient temperature Note that due to the Peltier element the difference in internal and ambient temperature is the key parameter 11 the minimum interior temperature that can be achieved is highly dependent on the external temperature If possible it would be preferable to be able to control the temperature assigning some initial value without sampling internal air or water temperature and make control decisions purely based on the time needed for cooling 16 Figure 2 8 shows the cooler air and water temperature over time as the cooler runs as well as the difference between those and the ambient temperature In this case the ambient temperature experiences relatively small fluctuations _ 60 ee water 40 cooler air 30 diff ambient int air a 20 diff ambient water o 0 50 100 150 200 time min Figure 2 8 Loaded Cooler Temperature Study Most intriguing is the linearity with which the water temperature decreases The downward rate of change of water temperature is approximately 05 F per minute This linearity is also observed in a warming scenario chilled water is placed inside the cooler w
7. F state cooler_load_c state on off condition 0 1 iTemp eTemp external temperature F gt eTemp Tdiff2 new temp difference F iTemp internal temp F Figure 4 12 Cooler S function Block The laptop load S function block will take as input the estimated initial SOC of the laptop lithium ion battery as well as the time under power and use a mathematical model derived from actual performance data to decide on the resistance of the load Port Identity t charge time m ate LSOC battery SOC initialization SD swH switch for high power load 0 1 LSOCout swL switch for low power load 0 1 LSOCout new battery SOC Figure 4 13 Laptop S function Block The charge time port t is driven by a running counter which is reset upon activation of the laptop load switch The laptop s battery management system charges the battery at a 54 nearly constant rate until it appears to begin to limit charge at around 80 capacity See chapter two for the laptop characterization The available Simulink packages at our disposal do not include a variable resistor Thus only two laptop current draw options are provided for the simulation low draw for a full battery and high draw for a non full battery In order to approximate the total power needed by the load over time the switch from high current draw to low should take place when the battery SOC is estimated at 90 At this stage
8. Figure 2 1 Photo of SuPER Cart Associated Loads 0 cee ceseeseeseeeeeseeeeecesecneeeeeeeaeenaes 7 Figure 2 2 Phase 0 Block Diagram lisis 8 Figure 2 3 SuPER Power Flow Diagram sessessssesssssessessesetsresesseseesseseesesseseesesseseesess 9 Figure 2 4 Circuit Breaker Industries CBI Breaker Response Curve ccccesees 11 Figure 2 5 Status System Interface Block Diagram 8 cccssseeseeeeceeeceeeneeeneeeneees 12 Figure 2 6 Pyranometer Data Careuitic cccescassavssedecceccsiaseusbunedssesstvadagvadesacesctsnissiaandeetadys 14 Fig re 2 7 Cooler Power Denda dat 16 Figure 2 8 Empty Cooler Temperature Study oooooocnoccnocccnococonoconononononnnoconocnnn conc cconncinos 17 Figure 2 9 Cooler 60 minute Cycle Temperature Stud Y coooonoocinnccnonocinccnonacannnonnnonnnonos 18 Figure 2 10 Cooler 725 minute Cycle Temperature Study ooooonncninconocncocccoonnconncconac ns 18 Figure 2 11 Lind Electronics Model DE2035 966 Converter c cooconcccocccoconoconaninncnnnens 20 Figure 2 12 Laptop Battery SOC Under AC Power coccoconcccocococonononnnonnconcconannonononnconnanns 20 Figure 2 13 Observed Laptop Current Draw Under Solar Powet esceeseeseeereeeeees 21 Figure 2 14 Lithium ion Battery Charging Current as a Function of Time 21 Figure 2 15 Motor Load Power vs Torque ccsscsccesscescesseceeeceecesceeseeseeceesenceesceseees 23 Fi
9. The broader focus of the work carried out on this thesis project is the effort to build a reliable self monitoring and adjusting 150W solar powered DC plant and distribution system In these early stages of development the loads considered are a small television electric cooler LEDs for lighting laptop and permanent magnet motor The first seven months of the SuPER team s efforts resulted in a partially complete open loop system dubbed Phase 0 The white paper mentions the goal of achieving a complete prototype system Phase 1 within one year of commencement A few months into project work the team felt confident in reaching and even exceeding that goal However the development of the DC DC converter a crucial subsystem hit a few road blocks Phase was not achieved by the end of 2006 as expected nor was it by March 2007 despite the fact that new converter teams came on board in October 2006 and February 2007 As aresult of these hardware setbacks we were inclined to turn our attention towards other efforts for the time being The software for the status and control system written in C was developed as necessary in preparation for the integration of the converter The SuPER team also recognized the knowledge that can be gained in simulating such a complex system in computer software and this thesis presents a complete first generation system model Such a simulation can reveal what types of load scenarios a 150W panel can support
10. calculates module current under given voltage irradiance and temperature Ta bp_sx150s Va G T 14 Out la Module operating current A vector or scalar In Va Module operating voltage V vector or scalar G Irradiance 1G 1000 W m 2 scalar Tac Module temperature in deg C scalar Written by Akihiro Oi 7 01 2005 Revised 7 18 2005 AE EE II II II II II IES Define constants double k 1 38le 23 Boltzmann s constant double q 1 602e 19 Electron charge Following constants are taken from the datasheet of PV module and curve fitting of I V character Use data for 1000W m 2 double n 1 62 Diode ideality factor n 1 ideal diode lt n lt 2 double Eg 1 12 Band gap energy 1 12eV Si 1 42 GaAs 1 5 CdTe 1 75 amorphous Si double Ns 72 of series connected cells BP SX150s 72 cells double TrK 298 Reference temperature 25C in Kelvin double Voc_TrK 43 5 Ns Voc open circuit voltage per cell temp TrK double Isc_TrK 4 75 Isc short circuit current per cell temp TrK double a 0 00065 Temperature coefficient of Isc 0 065 C Define variables double TaK 273 TaC 0 Module temperature in Kelvin double Vc Va Ns Cell voltage Calculate short circuit current for TaK e Ca Isc Isc_TrK 1 a TaK TrK lculate photon generated current given
11. DAO DAO DAO mxBaseRead mxBaseRead mxBaseRead mxBaseRead mxBaseRead mxBaseRead mxBaseRead mxBaseRead mxBaseRead AnalogF 64 BinaryI16 CounterF 64 CounterScalarF64 CounterScalarU32 CounterU32 DigitalScalarU32 DigitalU32 DigitalU8 Write Functions DAO DAO DAO DAO mxBaseWrit eAnalogF 64 mxBaseWri teDigitalU8 mxBaseWri teDigitalU32 mxBaseWri teDigitalScalarU32 Internal Buffer Configuration DAO mxBaseCfgInputBuf fer Error Handling DAO mxBaseGetExtendedErrorinfo 88 Appendix B Status Data Extraction Macro for Excel There are two versions of the macro one each for TIME FACTOR 15 and TIME FACTOR 30 They are both found in super_status_macros xls This macro performs a moving average with a window of width five and then downsamples the result by a factor NUM_ TO OUTPUT 60 Tss currently 150 to supply one sample per minute of run time Upon running the macro the data appears in columns S through AE All sensor data except the converter and battery temperatures are assessed The macro can easily be edited to include them if necessary 89 Appendix C PIC Serial Communication Protocol This protocol was originally created to facilitate testing over a Hyper Terminal interface but later adapted to the C code It was designed for the 16 series but also services the 18 series Type send M lt value gt to set
12. Decrease DC else if charge_mode 0 2 amp amp Vb 0 lt 12 7 reenter bulk mode must avoid thermal runaway charge_mode_out 0 1 DCnew DC O C Increase DC if DCnew lt 0 DCnew 0 if DCnew gt 1 0 DCnew 1 0 pi Update history DCprev 0 DC O DCout 0 DCnew Ppv 0 Pa_new D 3 Switch Control S function Code function swcontrol block function to control load switches based on scenario 1 1 all ton toff values are in minutes of the day Ts one minute in scenario number system time out switches 5 Ls G Written by Tyler Sheffield 2 7 07 ANA A AAA define INSTANCES 8 number of on off pairs for each load define INA 5000 defines inactive parameter never reached in time int 1 0 3 0 double table 5 INSTANCES set up time table double stimel000 stime 0 1000 int stime_int int stimel000 this skews the value for some reason tl gt ff 3 tv_ton ty toii cooler_ton cooler_toff light_ton light_totf laptop_ton laptop_toff re Ss motor_ton motor_toff if int scenario 0 0 table 0 0 INA table 0 INA table 0 2 INA table 0 3 INA table 0 4 INA table 0 5 INA table 0 6 INA table 0 7 INA table 0 INA table 1 INA table 2 INA table 3 INA table 4 INA
13. Pw Pb Poc P3 w BoB 8 BB time min Figure 5 18 System Power Levels 70W Load on CKT 3 78 Using this information we can capture an idea of the efficiency of the Outback MX60 converter at certain power input levels Figure 5 20 shows how the efficiency rises as the input power drops 110 105 e 100 95 Win ht HH E Di E 80 RA BO T 65 60 o 50 100 150 200 250 time min Figure 5 19 PV Power and Converter Efficiency In fact with input power of about 50W or less converter efficiency is near 100 Some of these phenomena uncovered while exploring status system information will certainly attract further investigation in the near future Averaging chunks of this data we can conjecture a bit as to the behavior of the elements involved in these losses assuming the given simple loss model see Table 5 1 Table 5 1 Estimates for Values of Loss Contributive Elements Characteristic PV Power Value MX60 efficiency 90W 92 MX60 efficiency 80W 94 MX60 efficiency lt 50W 100 Rl 90W 031 Q Rl 80W 029 Q Rl 50W 024 Q R2 90W 038 Q R2 80W 055 Q R2 50W 048 Q 79 There is a clear indication that R1 will increase as the PV array power increases The losses are greatly dependent on the amount of power being distributed as resistive losses ar
14. simulation bp_sx50 m bp_sx150s m PVmpp02 m script for writing all scope data out to Excel files writes to sim_waves last_sim_data xls storage_script m Tyson Den Herder s final model systeml3a mdl final SuPER model using C MEX S functions super_c mdl other stages in the model development process included in case of need as reference model_name mdl PIC Assembly pic Note that there are other required files that come packaged with MPLAB not supplied here The mplab folder contains the MPLAB setup files The diypack folder contains the MicroPro software The usbdrivers folder contains drivers for the programmer device old PIC 16F877A code pwm asm current PIC 18F4320 code pwm_18 asm compiled hex generated by MPLAB used to program PIC 104 pwm_18 hex MPLAB project file workspace SuperPWM_18 mcw other files needed by the project filet SuperPWM_18 Assembly code editing building and programming instructions Open MPLAB Click File gt Open Workspace Choose C Super SuperPWM_18 mcw Edit pwm_18 asm as needed Click Project gt Build All Connect the K128 programmer with PIC to the PC via a USB cable It may be necessary to check the COM port of the device with Device Manager Open MicroPro Click File gt Port and enter the COM port of the USB programmer Check that the Chip Selector is correctly set Click File gt Load Select the file C Super pwm_18 hex Click
15. 500 ms real time transient response Because of our need to speed up the simulation process system changes are made at much faster rates We therefore cannot get accurate results using the motor model as shown in the simulation of the model if we desire reasonable simulation times For broader scope system simulations we will replace the motor subsystem with a 6Q resistor representing a 237W load For an analysis of the transient response while the motor is in system we can adjust the sample times for the other subsystems in the model to allow the motor transient to proceed uninterrupted as it would on the prototype Figure 4 16 shows the simulated transient response 50 45 40 35 2 bat 25 lt Vbatt gt 20 15 10 5 o 7 7 7 7 o 0 1 0 2 0 3 0 4 0 5 time sec Figure 4 16 Motor Simulink Model Load Transient Witts will install a 58F ultracapacitor that will serve to protect the battery from deep current draw A version of the Simulink model with the ultracapacitor included was created and simulated As can be seen in the simulation results Figure 4 17 the battery current does not jump up to 40A but gradually increases it nears the 20A steady state level in about 30 seconds 57 Color Identity cyan battery current A magenta torque Ib in yellow battery voltage V red back EMF voltage V green speed krpm Figure 4 17 M
16. 8 highest order bits of the duty cycle register Type send L lt value gt to set 2 lowest order bits of the duty cycle register The PIC will echo back an exclamation point to confirm receipt The highest order bits range is integers 55 155 for a duty cycle of 0 100 e g an integer 65 will result in a DC of 10 There are only four possible values for the lowest order bits 0 3 Use chars 0 1 2 3 integers 48 49 50 51 to fine tune the duty cycle For example to get a duty cycle of 17 25 type MHLI in HT This will echo back as MH L1 To get a duty cycle of 45 75 type MdL3 This will echo back as Md L3 In C simply transmit four bytes M 72 L 49 for the first case or M 100 L 51 Any character entered that is not an M or L or not prefaced by M or L will be ignored and simply echo back followed by an 90 Appendix D C MEX S function Code D 1 PV Array S function Code function PV block wrapper S function around Aki s pv array model in G irradiance KW m 2 TaC temp deg C Vpv If out Ipv Adapted to C by Tyler Sheffield 2 14 06 EESAATA RARA MARA AA A TT AAT TT ATA TATTLE double Ia_new bp_sx150s Vs G TaC double Va Vpv 0 double ac 0 8 attenuation coefficient based on observed max power int j function la bp_sx150s Va G TaC function bp_sx150s m models the BP SX 150S PV module
17. PIC Microcontroller Load 2i Toggle and Le _ on charge oads wn Sensor C i anale I j j XS A e w 3 e PWM aN Serial a 2 Gs cycle and serial Interface TTL to il communication gt Serial This block diagram varies r ee al C Code Converter from the Phase 1 block D4 diagram in only two places 1 Outback MX 60 instead of Stage Integrate all DC DC converter individual system 2 Open loop PWM signal components to one unit on since there is no DC DC the cart converter to interface it with Figure 2 2 Phase 0 Block Diagram 2 The Phase 0 system uses the MX60 a high capacity DC DC converter manufactured by Outback Power Systems to buck the voltage level from the PV array to the desired 12V level on the load and battery bus The MX60 is not simply a converter but also an MPPT charge controller This feature has been critical for this early period of prototype development 2 1 1 Power and Distribution It was hoped that the Phase 0 system would provide about 400Wh of energy per day About half of that was earmarked for the 187W DC motor while the remainder would service the battery and all other loads However the motor s output rating is 187W The permanent magnet motor is not 100 efficient and to produce 187W of power the motor requires more than 240W of input power In addition the laptop needs to be running at all times throughout the day and requires at least 240Wh for an eight hour run As such
18. This first case illustrates the three hour nighttime lighting situation in which the laptop alone is run for one hour with the lights joining for the final two The data of Figure 5 5 is the battery voltage and current taken from the status system measurements Vb b voltage V 0 50 100 150 200 time from sunset min Figure 5 5 LED Light Two Hour Measurements The early aberrations are likely due to the laptop battery taking on a small amount of charge Figure 5 6 shows the results of simulating the same scenario 70 126 2 6 1258 28 12 56 3 2 T Vb g 12 54 3 2 5 3 4 E 3 12 52 g 3 6 12 5 3 8 12 48 4 o 50 100 150 200 time from sunset min Figure 5 6 LED Lights Two Hour Simulation The simulation predicts higher battery current draw and voltage The lights draw a relatively small current and a heavier load will be a more interesting case to examine The prototype was placed under test with the motor load on March 19 2007 a fairly sunny day with intermittent cloud cover Shown in Figure 5 7 is the insolation and temperature data for about five of the daylight hours 1000 30 ee 1 600 500 g 400 i E 300 2 200 100 0 r 7 7 7 0 328 378 428 478 528 578 time min G T insolation W m 2
19. an HTML report containing the coverage data generated during simulation of the model A large part of using model coverage is specifying model coverage reporting options in the Coverage Settings dialog box Some of the data generated by the coverage report includes total simulation time signal ranges and subsystem complexity details Note that activating coverage reporting may increase simulation time 100 Appendix H Improving Simulation Performance and Accuracy extracted from 31 Simulation performance and accuracy can be affected by many things including the model design and choice of configuration parameters The solvers handle most model simulations accurately and efficiently with their default parameter values However some models yield better results if you adjust solver parameters Design Factors in Simulation Speed Slow simulation speed can have many causes Here are a few When a model includes a MATLAB function block or M file S function the MATLAB interpreter is called at each time step drastically slowing down the simulation Using the math function block and C MEX file S functions will eliminate the need to invoke the interpreter Your model may include a Memory block Using a Memory block causes the variable order solvers ode15s and ode1 13 to be reset back to order 1 at each time step However this does not appear to be an issue when using the fixed step discrete solver The maximum step size may be too s
20. establish load time and power boundaries and provide information on how and when to best utilize battery energy storage while maximizing the life of the battery The simulation will also be critical for making plans for scaling the system up in size power The goal of the thesis is to show that a virtual mathematical model of the entire system compares favorably with a prototype system constructed entirely by students with faculty and staff guidance Simulink with its SimPowerSystems model package is used regularly in industry as a power and control system simulation tool and has been chosen for the SuPER simulation Achieving this goal will require characterization of the DC loads and careful study of prototype performance so as to allow proper modeling in Simulink As such the majority of the work done by the author during fall 2006 and winter 2007 quarters has been in these areas 1 constructing the Simulink model and simulating the system 2 developing the framework for the prototype status and control system software 3 testing the system operationally under a variety of load conditions 4 characterizing the loads 5 solving both hardware and software bugs that have periodically arisen 6 coordinating and supporting the undergraduate students working on senior projects associated with SuPER 1 5 Document Overview Chapter two introduces the SuPER system prototype as it existed in the fall of 2006 included is information about t
21. fairly consistent with the known charging current requirements for lithium ion batteries 13 Lithium ion batteries do not require a low current trickle or float charge and in fact may be damaged by such Thus the charge cutoff current is to be OA Figure 2 14 gives the shape of the expected lithium ion battery charge current over time charging current time Figure 2 14 Lithium ion Battery Charging Current as a Function of Time 21 For the vast majority of the operation time of the system the laptop will be a 30 35W load rather than a 54 60W load This issue will be addressed in simulation by providing a variable load controlled by a laptop specific function block The mechanism is present for future use but at this stage of development of the model the laptop battery is treated as always fully charged As the laptop is the intelligence of the entire SuPER prototype system it is necessary that it be powered throughout the operating period of the system It will thus be the last load to be disconnected from the system We will not be relying on the laptop s internal lithium ion battery for any sort of sustained operation of the status and control system at this time It will of course provide a small amount of power on current for the laptop when initiating system operation from a shut down state and will only be depended upon for that purpose 2 2 5 DC Motor For the SuPER prototype the team has equipped a 1 4 hors
22. h which is included in all source that interfaces to the NI DAQs Every time new interface code is written it should have these four accompanying folders and their contents Each src folder contains the source code executable and Makefile Simply type make at the prompt while inside the src directory to compile the source digpot src home super1 digpot src code for communicating over the 2 wire serial interface of the MAX5529 digital pot potcomm c cap src home superl cap src code for the switches that control charging and discharging the ultracapacitor cap c pvpro src home superl pvpro src the brain of SuPER where you will find maint contAcquireNChan c temporary storage file currently where the battery SOC is written Super_Output csv 103 pvpro pno home super1 pvpro pno PIC communication functions commpic c stand alone PNO code pno c PNO functions pnopal c MATLAB draft_model these are the original Level 2 M file S functions batt_voltage m control m control_plus m laptop_load m PV m swcontrol m files generated by MATLAB from the S function building blocks upon build command module_name _c c module_name _c mexw32 module_name _c_wrapper c module_name _c tlc the storage files for the C MEX S function C code not used by MATLAB module_name _src c files associated with Aki s PV array model and no longer directly used in
23. lt 0 discharge mode v1 0 1 95 18 B ns R1 0 19 1037 B 14 ns rd9 SOCm real else if I1 0 gt 0 charge mode v1 0 2 148 B ns R1 0 758 1309 1 06 B ns rc9 SOCm ee k V1 0 I1 0 SD SOC2 0 SOCm t phi9 soc SOC2 0 ee SOCm Soc2 0 SOC else if else if SOC1 0 gt 9 amp amp SOC1 0 lt 1 if 11 0 lt 0 discharge mode v1 0 1 95 18 B ns R1 0 19 1037 B 14 ns rd10 SOCm else af I1 0 gt 014 charge mode v1 0 2 148 B ns R1 0 758 1309 1 06 B ns rc10 SOCm ee k V1 0 I1 0 SD SOC2 0 SOCm t phil0 SOC SOC2 0 ee SOCm Soc2 0 SOC else if else if SOC1 0 gt 1 if 11 0 lt 0 discharge mode v1 0 1 95 18 B ns R1 0 19 1037 B 14 ns 50 SOCm else if 11 0 gt 0 charge mode v1 0 2 1 148 B ns R1 0 758 1309 1 06 B ns 30 SOCm ee k v1 0 11 0 SD SOC2 0 SOCm t 4 soc SOC2 0 ee SOCm Soc2 0 SOC else if Vbat 0 V1 0 11 0 R1 0 D 5 Laptop S function Code laptop_load_src c function laptop load block laptop load battery management mimic in time in minutes initial estimated battery SOC 0 1 out load select switches new SOC Written by Tyler Sheffield 1 15 07 AAA AAA AAA AAA AAA AAA AAA AAA AAA AAA A
24. made to factor temperature levels into model behavior 80 Chapter 6 Conclusion 6 1 Achievements The loop still has not been closed on the Phase 1 system as we set out to do in the fall of 2006 It was necessary to adjust some of the team goals to better fit the available human resources Focus was turned toward simulating the entire system in software After months of effort the first generation Simulink model presented herein has shown heartening results The SuPER team is now equipped with a viable first order Simulink model of the entire system Adjustable parameters will enable the simulation to provide greater service as a virtual representation of the SuPER prototype in the near future The model can then be easily modified to allow for increasing the PV array size introducing more efficient PV array technology introducing new battery models adding new loads better representation of power losses in the system heat etc development of adaptive control The status and control software structure is now ready for the future integration of the DC DC converter although some modifications may be necessary depending upon the manner in which the converter will be controlled The Phase 0 prototype has been put through its paces as all five developmental loads have been tested and characterized With the assistance of undergraduate students great progress on other system aspects such as LED lighting and high current
25. of development the laptop is always implemented as the lesser of these two loads as it is always considered to have a fully charged battery each morning 4 1 3 DC Motor Subsystem The design of the DC motor load proved to be an interesting problem Initially it was planned that Oi s Simulink motor model should be copied despite the fact that he was modeling a different motor than SuPER s However Witts was able to obtain more information about the motor parameters from the equipment manufacturer and in conjunction with experimental data was able to develop an accurate model for our motor in PSpice 16 This model was then ported to Simulink shown here as Figure 4 14 55 7MH 128 ohms Der MOM Vo Figure 4 14 Simulink Motor Subsystem This configuration uses current controlled voltage sources to represent back EMF and the torque of the load The torque is a constant 8 lb in Algebraic loops made unit delays necessary for the current measurements driving the voltage sources see section 4 2 for more on algebraic loops The simulation results are given as Figure 4 15 These results are achieved using an 11 75V constant source as a power source for the motor Color Identity cyan magenta yellow red green battery current A torque Ib in battery voltage V back EMF voltage V speed krpm Figure 4 15 Motor Transient in Simulink time in s 56 The motor has near to a
26. on the array and battery Note that in all of these examples the two figures share the same time index a insolation Wm 2 8 b Figure 5 14 Five Load Two Day Scenario One a Load Schedule b SOC Estimation 75 Figure 5 14b shows that in this case the battery SOC will clearly decline each day Perhaps the problem can be remedied by operating the motor only every other day Figure 5 15a gives the load schedule for a two day new scenario in which power needs are reduced to 1 583Wh for the two day period TV cooler lights A eee td O laptop motor a 1 05 1200 1 re 1000 lt T g 09 k r a soc 0 85 7 x 7 i 6 pees G a 400 Y 0 8 ri i i 5 0 75 Fi A 200 i 0 7 L E 0 o 500 1000 1500 2000 2500 time min b Figure 5 15 Five Load Two Day Scenario Two a Load Schedule b SOC Estimation Operating the motor only every other day results in a more sustainable operation scenario as the second day allows for some recovery for the battery however it can be seen Figure 5 16b that the SOC at the end of the second day is much lower than the initial SOC This is cause for concern if the motor is required to run on the third day 76 In this final case all cooler operation is halted and the motor is operated every day for one hour All other loa
27. pertinent to the success of SuPER but supplementary in nature to this thesis Chapter 2 Background 2 1 Phase 0 Prototype Tal spearheaded the effort throughout the first half of 2006 to assemble the first SuPER system The resulting system is a completely functional prototype and the current state of the system is known as the Phase 0 system Figure 2 1 SuPER consists of a PV array DC DC converter storage battery and DC loads Batteries are one of the most expensive components in the system as they cannot be manufactured on campus Two of the key loads in the system a water pump and small refrigerator are intended to be run primarily during hours of peak insolation but the SuPER team also considers evening lighting to be an essential load The improvement in safety and reliability of electrical lighting over fossil fuel consuming sources is worth the additional cost and complexity of including battery storage Figure 2 2 is the Phase 0 block diagram 77 Lo Loads MAX622 High Side Driver w MM74C903 Hex Buffer PVIT090 igh Sit 12 High Side v2 O v3 13 Ber 11 tL 72 EL T3 Oe TFT T l En PV Panel 150W Outback MX 60 Battery 12V Q i MO PWM Signal Open Loop Response PV out USB Interface PC A gt USB 6009 l gt r DEDE OUt Suin N
28. prototype 22 23 It has three primary charge states bulk absorb and float While in the absorb stage the MX60 gradually reduces current over time and assuming plenty of available current will run approximately one hour The battery documentation prescribes charging voltage levels for bulk and float stages Figure 2 16 but makes no mention of an absorb stage Table 2 1 Deka Battery Charge Voltage Guide 15 Temp Charge Float Temp F Optimum Maximum Optimum Maximum C gt 120 3 00 3 30 12 80 13 00 gt 49 110 120 13 20 13 50 12 90 13 20 44 48 100 109 13 30 13 60 13 00 13 30 38 43 90 99 13 40 13 70 13 10 13 40 32 37 80 89 13 50 13 80 13 20 13 50 27 31 70 79 13 70 14 00 13 40 13 70 21 26 60 69 13 85 14 15 13 55 13 85 16 20 50 59 14 00 14 30 13 70 14 00 10 15 40 49 14 20 14 50 13 90 14 20 lt 39 14 50 14 80 14 20 14 50 lt 4 25 The MX60 is programmed by the user with these voltage levels The absorb stage is entered immediately after the bulk charge stage Absorb and float stages are both employed only when the battery is in a high SOC The primary concern for battery integrity is avoiding overcharging which is the condition of supplying charging current when the battery is already at 100 SOC hence the different charging stages recommended by the manufacturer 15 For simplification of the SuPER status and control system
29. single components of the system and hence SuPER s anticipation of future breakthroughs on these technologies 6 3 Recommendations The need to finalize the Phase 1 system by completing integration of the DC DC converter cannot be overemphasized The Outback MX60 is strictly a temporary solution and much of the Simulink model s future effectiveness as a virtual system 83 modeling tool and test bed will depend upon the converter This is the single most important step for furthering SuPER progress The SuPER team may want to remove confusion by using the Celsius scale for all future cooler load work Fahrenheit has been used to this point because the manufacturer chose to describe the cooler characteristics on that scale The losses inherent in the prototype infrastructure ought to be investigated There is nothing that will restrict further progress on SuPER in this matter but achieving the highest possible efficiency may require a future redesign of the distribution bus side of the system The VRLA battery storage is another topic of interest whose characteristics may also contribute to some of the system losses however it is not entirely clear how much more effort should be expended toward properly modeling the battery Certainly some time should be spent towards including battery temperature as one of the model inputs but despite the large amounts of research done on these types of batteries they are still destined to be difficu
30. the key sample times in the model Table 4 2 Final Model Sample Times Entity Time s System Se 8 Insolation Temperature Data le 3 Control Block duty cycle 2e 4 Switch Control Block le 3 Battery 2e 4 Laptop le 3 Cooler le 3 PV array Se 8 Running Means le 4 The system requires a PWM signal generation block that can dynamically modify the duty cycle of the signal There is no such block in the Simulink library so it was necessary to create one the new subsystem is shown in Figure 4 20 The duty cycle value is used to alter the phase and amplitude of a sinusoidal signal oscillating at the switching rate The resulting sample is fed as input to a threshold based switch which produces either a zero or a step and alternates to form a square wave output This dynamically adjustable PWM signal generation subsystem is confirmed to operate equivalently to Simulink s PWM block 62 bias AE e Step dut Ind y Horiz Ca iit vax Digital Clock Concatenation sine Figure 4 20 Dynamically Adjustable PWM Signal Generation Unit The sine function block output expression is 2 7 t 2 2 D 50 25 Tw where Tsw is the inverse of the switching frequency in this case 2e D is the duty cycle value and t the simulation time The bias function block output expression is sin x 5 D One of the problems with the earlier versions of the system which included the PV array converter
31. the priorities of the Phase 1 control system needed to be reconsidered and this will be discussed in later chapters See Figure 2 3 for the Phase 0 1 power flow diagram Power consumed by sensor and switch boards lost in cable and switch resistances or otherwise unaccounted for and attributed to system losses is significant and preliminary loss investigations will be presented in chapter five PV 120 W w w2 8 wW Cooler sw3 60 W LEDs sw4 45W gt Laptop sw5 sow Motor sw6 250w IK Figure 2 3 SuPER Power Flow Diagram During previous work on SuPER the team had somehow overlooked the problematic issue of high current which we will call current above five amps traversing the copper traces on the switch board PCB This was not realized until motor testing one afternoon The load torque was increased to about 4 lb in which was higher than previously tested values At this level the motor seeks to draw upwards of seven amps perhaps eight or nine depending on the load bus voltage It turns out that the PCB traces were thick enough only for about five amps and the increased current caused the traces to heat up and melt the solder joints at the MOSFET This in turn created a short at the MOSFET terminals The problem was first noticed when the status system reported a draw of 39A from the battery due to the newly created short which would be possible in a scenario with many running loads b
32. these particular blocks they are known as Level II M file S functions The function that is to be implemented can be written in MATLAB language just as would be done for execution or function call from the MATLAB command line However creating an S function requires wrapper code around that function code Ports must be enumerated and identified for each input and output of the block The function code is placed in a separate section for determination of the outputs One drawback of these S functions is their lack of internal memory In other words the function is executed top to bottom continually with no memory of previous states Efforts to speed up the simulation process led to a new approach to the S functions The software is greatly hampered by the need to call the M interpreter each time M file S functions are invoked By writing the code in C instead and pre compiling 1t before runtime the simulation time can be greatly reduced This variety of function block is known as a C MEX S function Running one simulation on a 1 4 GHz machine for the M code blocks required 22 hours The same simulation on the same machine with the new C code blocks runs in under two hours Insolation and temperature data are located in lookup tables LUTs as described in Den Herder s senior project 20 We will continue to use Oi s model for the PV 46 array as defined in his thesis paper 28 Figure 4 5 is a diagram of this model which was implemen
33. to my pool of engineering resources and tools 1 3 Solar Power Systems In section 2 1 of Tal s 2006 thesis paper 2 he convincingly outlines the case for SuPER only a summary of his arguments will be presented here There are multitudinous opinions on whether or not rising global temperatures are directly caused by human activity regardless of the cause it is nevertheless a fact that atmospheric carbon dioxide levels and by extension temperatures have been sharply on the rise in the last 25 years 3 Such changes will have consequences for life on this planet as we know it SuPER harvests energy from a renewable source and contributes no direct air pollution to the environment It is a device designed with the goal of sustainability in mind It is also intended to be a low cost system which will pay for itself within a short time of activation 1 in order to provide advantages to lower income families who have not previously had access to a power generation system There is no grid infrastructure required as all issues associated with long distance power distribution are removed as costs obstacles and energy sinks Solar cell technology is becoming increasingly important as an energy source for reasons alluded to above As a result it is also becoming a more ubiquitous better researched more efficient and more cost effective technology 4 The technology is quickly developing into a preferable option among those in SuP
34. we will abandon the absorb stage and utilize only bulk and float charging In discussions with Tal he indicated that this was a reasonable simplification 2 4 Phase 1 The closed loop version of the system in which all modules besides the PV array battery and laptop hardware are designed and created at Cal Poly is called the Phase 1 system Figure 2 17 is the Phase 1 block diagram The SuPER team s goal was to complete this phase by March 2007 Critical to success in reaching Phase 1 is the implementation of a locally designed and built DC DC buck converter specific to this application The MPPT control would then be moved to the status and control PC Two previous efforts at constructing a functional converter have already been made and did not succeed Perhaps the SuPER team had underestimated the difficulty in implementing such a device for this high current application Seniors Robert Casanova and Joe Shein began new efforts to develop the converter in fall of 2006 A second effort by seniors Thaddeus Guno Koosh Shah and Kunal Shah using an alternate approach commenced in early 2007 26 DAL Y i Loads MAX622 High Side Driver w MM74C903 Hex Buffer PvI1090 n a n igh Si ADA D High Sid v2 12 v3 13 Driver j 12 re m m ky a PV Panel 150W i DC DC Converter f gt Battery 12V vit PWM Signal PV Out y U
35. which SuPER uses is a 12V DC black white GPX portable TV radio equipped with a 5 inch screen It draws a continuous current of between 600 700mA 8W The television circuit is identified on the prototype as circuit 2 2 2 2 Cooler The Coleman 12V DC cooler was chosen to represent a typical low power 60 70W refrigeration device that might be used by families who have had no previous access to in home refrigeration The cooler load is identified as circuit 3 This particular model 5644 has a volume of 40 quarts and uses a Peltier element to cool the interior down to about 40 F below ambient temperature An empty cooler reaches this state in three hours The power cable is equipped with a 7 5A fuse For SuPER we wish to study methods of limiting the power needed in operating the cooler 15 It was hoped that the cooler would require less power to maintain a minimum temperature than would be needed to reach it Tests done with an empty cooler have shown this to be true Figure 2 7 illustrates the decrease in power consumption over time for the cooler 80 50 70 45 4 60 50 S y 30 Power 40 ac deltaT F o 30 4 20 a 5 145 5 e 20 10 10 5 0 7 7 7 0 10 60 110 160 time min Figure 2 7 Cooler Power Demand However an empty cooler being largely worthless we choose to characterize it while under a
36. 3 June 1998 Vasquez Gustavo Data Acquisition and Sensor Circuits for the SuPER Project Senior Project report California Polytechnic State University 2006 Apogee Instruments Inc Silicon Pyranometer Specifications Online Retailer lt http www apogee inst com pyr_spec htm gt Frohlich Claus Construction of a Composite Total Solar Irradiance TSI Time Series from 1978 to Present PMOD World Radiation Center May 2006 Belov I M M P Volkov and S M Manyakin Optimization of Peltier Thermocouple Using Distributed Peltier Effect 18 International Conference on Thermoelectrics 1998 Zukowski Joey Senior Project report California Polytechnic State University 2007 85 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 Allbatteries UK Ltd Charging Lithium ion Batteries Online Retailer lt http www powerpacks uk com Charging 20Li 10n 20Batteries htm gt Cao Jennifer SuPER Project Wiring and Protection System Senior Project report California Polytechnic State University 2006 East Penn Manufacturing Inc Valve Regulated Lead Acid Technical Manual Technical Manual 2004 Witts Joseph Using Ultra Capacitors for Energy Storage in Cal Poly s SuPER Project Senior Project report California Polytechnic State University 2007 Vairamohan Baskar State of Charge Estimation for Batteries Master s Thesis
37. AA AAA AAA AAA AAA AAA AAA AAA LSOCout 0 LSOC 0 11 new SOC calculation kel mexPrintf lf n t 0 if LSOCout 0 gt 1 LSOCout 0 1 else calculation of LSOC here if LSOCout 0 gt 9 swH 0 0 93 swL 0 1 else if LSOCout 0 lt 9 swH 0 1 swL 0 0 D 6 Cooler S function Code cooler_load_src c function cooler_load block cooler temperature calculation works only for nearly constant ext temp in initial interior temperature power state external temp out new internal temp Written by Tyler Sheffield 3 19 07 et AAA AAA AAA AAA AAA AAA AAA AAA AAA AAA AAA AAA AAA AAA AAA set to one minute equivalent sample time 8 quarts water values double w_rate 008 warming rate deg min double c_rate 05 cooling rate deg min double Tdiff 0 eTemp 0 iTemp1 0 if state 0 0 iTemp2 0 iTemp1 0 w_rate else if state 0 1 iTemp2 0 iTemp1 0 c_rate Tdiff 0 eTemp 0 iTemp2 0 96 Appendix E Choosing a Fixed Step Solver extracted from 31 When the Type control of the Solver configuration pane is set to fixed step the configuration pane s Solver control allows you to choose one of the set of fixed step solvers that Simulink provides The set of fixed step solvers comprises two types of solvers discrete and continuous The fixed step discrete solver computes the time of the next time ste
38. ATT and BUS Execute the software with the command contAcquireNChan Flip the breakers as desired to power indicated loads To shut down the software use q 0 Shut everything down by opening all circuits at the breakers 102 Appendix J File README SuPER Control System and Simulation README Author Tyler Sheffield tyler galatix com 760 460 6880 3 28 07 Last Update 4 25 07 The files described in this readme are organized by directory local to this CD File names have type extensions while descriptions are bookended by Directories on Linux are given in parentheses if applicable Obsolete files no longer used by the sim or master control are marked with a Name placeholders or wildcards if you will are between An is a file type wildcard A General Files a collection of observations and important things to remember compiled during SuPER s lifetime SuPER Master Documentation doc the final thesis doc sheffield_thesis_4 25 doc the defense power point presentation SuPER Defense Presentation ppt SuPER lab bulletin poster files nov 2 presentation color rev ppt nov 2 presentation ppt nov 2nd presentation alt 7 pdf Laptop C Code c_code All project code is found on the laptop in the directory file home superl The pvpro bin pvpro etc and pvpro lib folders contain various utilities for the NI DAQs to work properly The pvpro include folder has the file NIDAQmxBase
39. C SS 40 C sc 104 F 46 F STORAGE TIME MONTHS wW al S gt E 2 a q o a E q a ta 27 Figure 4 9 Deka VRLA Battery Self discharge Chart 15 This information yields a discharge constant of 4 34e 5 h 5 11520 hours Another critical battery parameter is SOCm the total energy capacity in Wh This value is a function of the current draw and follows a somewhat logarithmic curve Seven current capacity data points are available in 30 only for the absorbed glass mat AGM version of SuPER s gel 8G31DT For the gel variety we know only that the capacity is 97 6Wh 4 88A which is slightly less than the AGM battery In consequence we must estimate what the gel battery curve may look like starting from the one known point Figure 4 10 shows the provided AGM curve and the estimated gel curve 50 1400 1200 800 e ACM 600 gel est capacity Wh 0 10 20 30 40 50 60 70 80 current A Figure 4 10 Current vs Capacity for AGM and Gel Batteries It was necessary to make adjustments by adding code to estimate the capacity SOC for different battery currents I in Simulink equivalent to the model s I For currents in excess of 2A we will use the logarithmic function derived from curve matching in Excel to make the estimate SOC 179 68 In 1 1435 2 For currents less than 2A a value of 1325Wh is use
40. Cal Poly SuPER System Simulink Model and Status and Control System A Thesis Presented to the Faculty of California Polytechnic State University San Luis Obispo In Partial Fulfillment of the Requirements for the Degree of Master of Science in Electrical Engineering by Tyler Sheffield April 2007 Authorization for Reproduction of Master s Thesis I grant permission for the reproduction of this thesis in its entirety or any of its parts without further authorization from me 11 Approval Title Cal Poly SuPER System Simulink Model and Status and Control System Author Tyler Sheffield Date Submitted 25 April 2007 Dr James Harris Committee Chair Signature Dr Ali Shaban Committee Member Signature Dr Jim Widmann Committee Member Signature iii Abstract Cal Poly SuPER System Simulink Model and Status and Control System Tyler Sheffield The Cal Poly Sustainable Power for Electrical Resources SuPER project is developing a solar power DC distribution system designed to intelligently service almost any load that might be needed by a single off grid household A prototype has been constructed and tested This thesis describes the creation of a modular MATLAB Simulink model of the entire system whose principal components include a PV array DC DC buck converter lead acid battery various loads and a digital status and control subsystem Also presented is the design of the status and control software
41. ER s target market where cooking heating and lighting energy needs are largely still provided by fossil fuels 5 There are a few commercially available solar power systems similar in scope to SuPER such as those manufactured by SunWize www sunwize com SuPER is an attempt to develop one of these types of systems at much lower cost and the team anticipates future advancements in technology that will make this possible This is especially true of solar cell and battery technology What is unique about SuPER is how it is put together and perhaps more importantly why In 6 Sharaf and Ul Haque present a DC motor solar power system along with Simulink models but there is no storage in the system In 7 by Chiang Chang and Yen a system very similar to SuPER is proposed and prototyped although it is designed to be a supplement to grid power rather than a replacement This is the case with many commercial systems For related reasons the authors are unconcerned with managing individual loads and optimizing battery life There are countless additional published papers that address a wide variety of other issues with the components that make up SuPER including but not limited to converter topologies battery state of charge SOC measurement and maximum power point tracking MPPT techniques Most publications referenced by the SuPER team do not propose or demonstrate a system on the scale of the SuPER project 1 4 Thesis Objectives
42. Fuses and make sure the Oscillator pull down menu is set to HS Click Program Excel macros status system data extraction macros see thesis Appendix B for usage instructions super_status_macros xls Data Files prototype_data status system data organized by date of operation sim_waves this folder is full of a variety of stored simulation results just a couple groups are identified here images of sim waves from first successful 24 hour period complete simulation run ran for 22 hours real time first all name bmp simulation results of insolation and temp data taken from prototype motor operation periods march date motor xls All Others drawings MicroCap schematic drawings drawing name cir Visio drawings flowcharts etc drawing name vsd experimental_data A repository for all other data and worksheets etc that had no place anywhere elses 105
43. PCB manufacturing has been made These 81 simulation tools and other development processes that have been established will facilitate the achievement of SuPER s goals for the next few years We are well on our way 6 2 Reflection on System Sensitivities Despite some hardware setbacks the SuPER project has arrived at a key point in the development process Many of the limitations of the design have been uncovered and assessed and work identified for years ahead The battery turns out to be one of the most difficult problems because of the complexities involved in accurately modeling it In order to create a system that will be cost effective we desire the battery to last as long as possible To create control software that will optimize battery life intimate knowledge of the battery s characteristics must be obtained There are other battery technologies that are more reliable and more easily characterized than the VRLA variety but suffer from other limitations such as stunted storage capacity Improvements in electrical energy storage technologies will hopefully usher SuPER towards greater viability Besides optimizing battery life there are two fundamentally important challenges that will confront the future generations of SuPER project collaborators lowering system cost and finding ways to utilize power more efficiently SuPER s success will of course rely on advances in PV cell technology as well The goal of the SuPER project is
44. SB Interface PC O USB6O0 me DC DC Out AN a SN PIC Microcontroller Pade Toggle and control J PWM duty i Serial va VL cycle and serial Interface TTL to 2 Ss communication e Serial C Code Converter Sensor C algorithm C 4 Code code from j E MATI 48 12 Notes Qi Q2 1 All loads and load probes j y are represented as one in this Qu diagram A 2 All probes are connected y to the USB 6009 via an op Stage Integrate all amp gain circuit omitted individual system from this block diagram components to one unit on 3 The combiner box which the cart doesn t appear in the block diagram junctions all the power lines Figure 2 16 Phase 1 Block Diagram 2 The specifications given for the converter derived from PV array and battery characteristics are Table 2 2 DC DC Converter Specifications 2 Parameter Value Input voltage wide range 0 to 40V Input current 4 75A max Max power 150W 80 efficiency target Output voltage 11 5 14V Output current 13A max Switching frequency 500 kHz 27 See section 5 2 2 in 2 for more details on the specifications Guno Shah and Shah are designing a converter from the ground up 24 and their efforts will include layout of a high current PCB The 500 kHz PWM signal for this converter is sourced by a microcontroller and some of the difficu
45. University of Tennesee 2002 Duryea Shane Syed Islam and William Lawrence A Battery Management System for Stand Alone Photovoltaic Energy Systems IEEE Industrial Applications Magazine May June 2001 Castaner Luis and Santiago Silvestre Modeling Photovoltaic Systems John Wiley amp Sons Ltd 2002 Den Herder Tyson Design and Simulation of Photovoltaic SuPER System Using Simulink Senior Project report California Polytechnic State University 2006 Fasih Ahmed Modeling and Fault Diagnosis of Automotive Lead Acid Batteries Master s thesis The Ohio State University 2006 Outback Power Systems Inc MX60 PV MPPT Charge Controller Installation and User s Manual 2005 Outback Power Systems Inc MX60 Specifications Datasheet 2005 Guno Thaddeus Koosh Shah and Kunal Shah Senior Project report California Polytechnic State University 2007 Casanova Robert and Joe Shein SuPER DC DC Buck Converter Senior Project report California Polytechnic State University 2007 National Instruments Corp User Guide and Specifications USB 6008 6009 Technical Manual 2005 National Instruments Corp NJ DAQmx Base 2 x C Function Reference Help Technical Manual 2005 86 28 29 30 31 Oi Aki Design and Simulation of Photovoltaic Water Pumping System Master s thesis California Polytechnic State University 2005 Taufik A Crash Course in Switching Regulators Sli
46. and simple battery were algebraic loop errors These result from the necessity of feeding back certain values into the function blocks There were particular difficulties with the PV block the source voltage V Vpv which is determined in part by the PV array current output also serves as an input to the PV block for use in calculating the current Simulink s help files declare An algebraic loop generally occurs when an input port with direct feedthrough is driven by the output of the same block either directly or by a feedback path through other blocks with direct feedthrough 31 In many cases Simulink has the ability to successfully navigate through the algebraic loops However one particularly insidious problem was a simulation halting loop calculation 63 error that would occur partway through a run and could not be foreseen The only known solution is to eliminate all loops completely with a work around adding a delay on the feedback path in the form of a 1 z Unit Delay block The amount of delay is one sample of the system sample time This much delay will not adversely affect the S function block output calculations as it is negligible in comparison to the clocking rate of any blocks Once the simulation was able to proceed without encountering loop errors it was found that the simulation strained system memory resources The amount of data needing to be recorded overwhelmed our machines We attempted to alleviate the pr
47. as remedied by adding code to alter the execution shell to return characters byte by byte from stdin with each key press When a q is pressed the loop catches it and is able to stop and clear all active tasks before halting the process See Table 3 1 for details on the USB 6009 device errors encountered by the SuPER team 32 Table 3 1 Known USB 6009 Errors Error Implication Action Required Shell message Device Linux has lost track of the Hard reset of DAQ devices identifier invalid DAQ devices Shell message Physical channel specified does not exist on this device No known cause Hard reset of DAQ devices Shell message Onboard device memory overflow Host processes have taken away system resources from the USB to PC data transfer or less likely the sample rate is too high Close all other executables and do not run anything besides status and control program Sensor readings are bogus such as large negative The device identifiers have been mixed up Hard reset of DAQ devices temperature values 3 2 Functional Overview All initialization and parameterization of NI DAQ tasks is handled in function main of the SuPER code which is found in contAcquireNChan c The code enters an unterminated while loop that repeatedly reads the values out of the storage buffers recorded by the USB 6009 and displays them onscreen Thousands of values are loaded by the USB 6009 into the b
48. ccsssceessecsseceeceeeeeescecsaeceseeseceesaeecsaeceeenees 33 Table 4 1 Battery Model Parametros 49 Table 4 2 Final Model Sample TM da 62 Table 5 1 Estimates for Values of Loss Contributive Elements ooooonnccnnnninncniococonccnno 79 Table 6 1 SuPER Development Costs to Date ooooonnoccnonncinnonncconccconoconanonncconccconocinos 83 ix Chapter 1 Introduction 1 1 The SuPER Project The Sustainable Power for Electrical Resources project was born in July of 2005 with Dr James Harris white paper describing a durable low cost family owned solar power system 1 It is intended as a self contained and self monitoring off grid DC system with energy storage capability that will service a wide variety of loads It was anticipated that development time of the system would be around five years with the first three years dedicated to research design and the building of a prototype system A goal of SuPER is to demonstrate that the system can extend component life especially that of the battery and achieve very low failure rate It is expected that SuPER will be used by family units in low income high insolation areas of the world 1 2 Personal Involvement I first learned about the SuPER project at a presentation made by Dr James Harris at one of the Friday afternoon sessions of the department s weekly graduate student seminars Photovoltaic cells are a fascinating technology and though I knew very little about p
49. d The battery S function is shown below in Figure 4 11 Port Identity Il battery current A same as Ie SOCI SOC initialization SOC2 new SOC Vbat battery voltage V Vi internal voltage V for debug RI internal resistance Q for debug Figure 4 11 Battery S function Block The battery model requires an initial SOC and some sort of state memory to properly update the battery condition throughout the simulation runtime The SOC is fed back 51 into the battery model input through memory blocks so that previous outputs will be held on the signal line The creation of the battery S function is followed by the adaptation of the model to the performance characteristics of the battery in use At our disposal is the battery charge and discharge test data acquired under operational conditions by Tal and published in 2 It is essential to note that the authors in 19 state that the battery model presented is only accurate for a SOC range of 30 80 Model adjustments can then be made experimentally in an attempt to match the data recorded while the battery SOC was in this range Castaner and Silvestre provide an example of adjusting parameters to fit a commercial battery but unfortunately do not explain their methodology Doing our best to follow their lead the same parameters will be altered Exactness in the battery model is not our chief concern but some precision for the 80 100 char
50. de Presentation Notes 2006 East Penn Manufacturing Inc Absorbed Glass Mat Series Datasheet 2003 The MathWorks Inc MATLAB User Guide Technical Manual 2005 87 Appendix A NI DAQmxBase 2 1 API Function List Function Purpose Function Names Task Configuration Control DAQ DAQ DAQ DAQ DAQ DAQ DAQ mxBaseClea mxBaseCrea mxBaselsTa mxBaseLoad mxBaseRese mxBaseStar mxBasestop rTask teTask skDone Task tDevice tTask Task Create Analog Input Channels DAO DAO mxBaseCrea mxBaseCrea teATThrmcplChan teAIVoltageChan Create Analog Output Chan nel DAO mxBaseCrea teAOVoltageChan Create Digital Input Chan nels DAO mxBaseCrea teDIChan Create Digital Output Cha nnels DAO mxBaseCrea teDOChan Create Counter Input Chan nels DAO DAO DAO mxBaseCrea mxBaseCrea mxBaseCrea teCIPeriodChan teCICountEdgesChan teCIPulseWidthChan Create Counter Output Cha nnels DAO mxBaseCrea teCOPulseChanFreq Timing DAO DAO mxBaseCfgsS mxBaseCfgI ampClkTiming mplicitTiming Triggering DAO DAO DAO mxBaseDisa mxBaseCfgD mxBaseCfgA bleStartTrig igEdgeStartTrig nlgEdgeStartTrig Reference Trigger DAO DAO DAO mxBaseCfgA mxBaseCfgD mxBaseDisa g nlgEdgeRefTrig igEdgeRefTrig bleRefTrig Read Functions DAO DAO DAO DAO DAO DAO
51. designed for PSpice but modified by team member Tyson Den Herder for Simulink as his senior project 20 This model is divided into charge and discharge mode and provides a SOC estimation using the battery energy balance equation Upon integration of the student designed DC DC converter it is anticipated that this model will be ported to C code on the status and control laptop In 21 Fasih provides some vindication of the model and methodology Like SuPER Fasih also made use of Hall effect current sensors and NI DAQ hardware for his measurements 24 One shortcoming of the model in 19 is that it disregards the non trivial impact of temperature on the battery state Also Castaner and Silvestre mention that the model realistically should be restricted to use for a SOC within a range of 30 80 of capacity for which it will provide the best estimates This provides a problem for SuPER as we desire to always maintain as high a SOC as possible The approach taken to this matter is detailed in chapter four Tal s thesis 2 Appendix B discusses the way battery charging is handled by the Outback MX60 converter The device relies solely on PV and battery voltages for its calculations as we will do with the Phase 1 system Our goal will be to mimic the operation of the MX60 with our converter The MX60 is a very expensive and efficient piece of hardware that was designed to handle much larger amounts of power than that associated with the SuPER
52. ds are unchanged with the needed energy now reduced to 1 399Wh See Figure 5 17 for the schedule and results SN a cooler lights laptop ie AN motor a b Figure 5 16 Four Load Two Day Scenario Three a Load Schedule b SOC Estimation The battery is now able to build charge on sunny days which will allow some reasonable DOD to occur on cloudy days without significant repercussions to battery lifetime 71 5 4 Power Losses Systems whose characteristics include high currents traversing non trivial distances face losses due to resistances in components such as cables and switches SuPER is no exception Model development has not reached the point where sophisticated representations for these sinks have been developed Using the measured voltage drop and current flow between the converter and battery and between battery and loads some idea of losses can be estimated Consider a simple resistive loss model Figure 5 18 where R1 and R2 represent switch and cable resistances between subsystems R1 R2 DC DC BATT E Figure 5 17 Simple Resistive Loss Model Some data has been gathered on system losses under multiple load scenarios Figure 5 19 shows power levels for various system components while running a 70W load for 250 minutes P3 represents the power consumed by the load 120 100
53. e proportional to the square of the current R2 is more difficult to qualify The 10 gauge wire connecting the DC DC converter output to the battery and loads runs approximately twelve feet in length twelve feet of 10 gauge wire offers about 02 Q of resistance Presupposing that we can attribute some of the measured losses to the wiring but not all this value seems to corroborate what has been observed This first generation Simulink model has had a collective 04Q resistance introduced on the converter and battery output a value chosen to lie between the calculated wire resistance and the estimated simple loss model resistances representative of the uncertainty in the true sources of these losses Future work on SuPER will need to examine these effects Analysis should be made towards the end of discovering where exactly these losses occur and how they relate to the voltage and current Potentially the highest sources of loss may be the cables battery inefficiency and switches Also the various sensor boards will consume some power A possibility worth investigating would be increasing the voltage and decreasing the current on the load side of the converter perhaps by going to a 24V battery Another factor for inconsistency to keep in mind is that only the PV array block takes temperature into account for calculating power output In reality the battery and converter subsystems will also be dependent upon the temperature Improvements can be
54. e the new battery SOC and the loop was used to dictate a step size and rate of SOC updates This allows the user to start with a given SOC and find the resulting SOC after any period of time with the battery under some known constant current flow The for loop was removed for the new simulation as the steady state current will be changing at a known rate equivalent to the highest frequency clocking found in the model most likely the control block Since the current may and likely will change with each duty cycle adjustment we desire to update the battery conditions at the same frequency We can thus fix the integration time window to be the sample time of the highest frequency clock in the model which will necessarily be the same sample time of the battery function 53 The work done in characterizing the various loads under operating conditions allows the implementation of some of the loads as S function blocks containing code that reflects actual power demands and responses In the case of the cooler load given the operating parameters described in chapter two the load will be represented only by a resistor in this first generation of the model However a function block Figure 4 12 was constructed to perform the temperature adjustments constantly occurring inside the cooler currently it is only effective for slowly changing external temperatures gt Taiff1 Port Identity Late P Tdiffl initial temp difference
55. eanasaesaectoneesanays 54 Figure 4 14 Simulink Motor Subsystem cc cecesseeseeeceeeeceseceseceeeeeeeeseeeaeceeeeaeeeneeeeees 56 Figure 4 15 Motor Transient in Simulink time in ooooonnccnnoccnonocononcnnnoconocananonnncnnnonns 56 Figure 4 16 Motor Simulink Model Load Transient 00 0 ceeeeseeseeneeeeeceeeeeeeneeeeeeeeees 57 Figure 4 17 Motor Model Simulation with Parallel 58F Capacitor time in s 58 Figure 4 18 PWM Signal Sampling a 5 Duty Cycle b 9 c 10 d 1 60 Figure 4 19 Discretized Converter Transient Response time in MS cccceseeeeteees 61 Figure 4 20 Dynamically Adjustable PWM Signal Generation Unit eee eeeeeeeeeees 63 Figure 5 1 Golden Colorado Insolation and Temperature ooooocnnccnocnnocconccononononanonncnnnons 66 Figure 5 2 San Luis Obispo Insolation and Temperature oooonocnnncnocanocconcnonononcnnonncnnnens 67 vil Figure 5 3 Nighttime LED Operation Simulation 0 ceceeccesecsceeneeeeeeeeeeeeeeeeeeeeeees Figure 5 4 Motor Operation SIA rinda did Alai Figure 5 5 LED Light Two Hour Measurement 0 c cceccesseeseeereeeeeeeecesecneeeeeeneeeeees Figure 5 6 LED Lights Two Hour SimulatiON coonoonioninnnoononnnnnncononan cono noronnanonnnnncrnonnns Figure 5 7 March 19 2007 SLO Insolation and Temperature cocicninnonnncnnniocinncnnsss Figure 5 8 March 19 2007 Motor Operation Measurements ccioicnninninncinocioc
56. ents It is probable that the arc in the battery current while the motor was running is due to the dynamometer torque unexplainably creeping downwards Figure 5 13 shows the same scenario in simulation 20 5 10 A AAA a 5273 323 373 423 473 523 573 623 lo 10 15 p 20 25 current A voltage V al time min Figure 5 13 March 29 2007 Motor Simulation Again the battery Voc was found the next day to indicate a fully charged battery We can only conclude that when starting motor operation with a full battery there is plenty of available solar power to recharge the battery promptly whether the motor is powered an hour before peak insolation or during peak insolation Future studies will adjust the 74 battery model accordingly and may then take into account cloudy periods that force the battery to begin succeeding days with less than a full charge 5 3 Multi Load Scenarios With the simulation we can consider a variety of load scenarios over lengthy periods of time and view the ultimate effect on the battery In this first example we run all five loads every day television two hours cooler 1 67 hours lights three hours laptop 14 hours and motor one hour Again the summer insolation data from Colorado will be utilized Figure 5 14a shows the load activation schedule for two days which places a demand of 1 833Wh
57. epower 12V permanent magnet DC motor which will be used to represent the water pump load It is anticipated that this is the most demanding load to be powered by SuPER Jennifer Cao s senior project report 14 records operational data for the motor also echoed here in Figure 2 15 22 250 200 m S leAt e power in a power out 100 Dar efficiency 50 eet E a o LT o 2 4 6 8 10 torque Ib in Figure 2 15 Motor Load Power vs Torque For the SuPER prototype testing we will plan on operating the motor at a constant 8lb in torque which loads the motor to near the maximum rated power output This is also a more efficient use of input power than operating at a lower torque A dynamometer is used to load the motor On starting up the motor there is a large amount of current drawn for a very short time period known as the inrush current and is accompanied by a correspondingly large drop in battery voltage Frequent repetitions of such current draw can have adverse effects on battery life over time if the battery charge is not maintained at a high level 15 and for this reason senior Joe Witts explores the advantages of including an ultracapacitor in the system for his senior project 16 He reports that such a configuration provides negligible assistance in the case of infrequent motor activation Cycled motor use with a
58. essential however to find a type of lighting that provides high output usually measured in lumens at minimum energy use To this end SuPER will rely on the emerging LED lighting industry The LEDs available today use about 3W of power and generate around 100 lumens each 12 Unfortunately operating at this wattage is inefficient The SuPER prototype will instead operate the lights at a tad over 1W apiece Four such LEDs will be allocated for use for the immediate future requiring approximately 4 5W of power The lights are designated as circuit 4 2 2 4 Laptop System status and control is run on a Dell Inspiron B120 laptop chosen for its low cost under 500 Specifications for the device declare it to be a 60W max system drawing about 3A at about 20V A converter is thus required for the 12V SuPER system 19 bus to which the laptop will connect A Lind Electronics Model DE2035 966 converter was purchased from Dell this converter will turn a 12 32V DC input into a 20V DC output at a maximum of 3 5A It is equipped with a 15A fuse Figure 2 11 shows the FB Figure 2 11 Lind Electronics Model DE2035 966 Converter converter Information about the power and battery management system on Dell s laptops is proprietary and therefore not available to the public However regular observation of the system in operation reveals some useful trends The approximate battery SOC while charging from a wall outlet is recorded u
59. for a particular model thus generally requires experimentation 98 Appendix F Managing Scope Data In the scope parameters dialogue box click on the data history tab Check the save data to workspace box and type the name of the struct wherein to store the data At the format menu select Structure with time When the simulation has finished running your data will be available in the MATLAB workspace identified by the previously specified name Double click on the desired struct to open the array editor To display the recorded scope data in a spreadsheet form perform the following actions double click the signals box double click the cell in the column corresponding to the desired signal double click the values box The data can be plotted by selecting the column and clicking on the plot icon of the Array Editor toolbar To export the data to Excel run the m file storage_script m it will take a moment for the data to be exported The data is written to a file in the MATLAB sim_waves directory called last_sim_data xls Setting the Scope Decimation Value Open the scope window and click on the parameters button in the upper left On the bottom is a text box marked Decimation Enter the decimation value in the box 99 Appendix G Specifying Coverage Report Settings extracted from 31 Coverage report settings appear in the Coverage Settings dialog accessed through the Tools menu Select the Generate HTML Report option to create
60. ge range would be beneficial as we hope to consistently maintain a high SOC on the battery The team adopted what is the perhaps the only reasonable approach to this issue which is to use Tal s data to make adjustments to the model in use for an alternate high SOC case The most difficult decision to make is how to model the battery in the tricky 80 90 charge range This is just above the model s effectiveness range and according to Tal still within the bulk charge mode breadth for the MX60 which extends to approximately a 90 charge The model is divided into four SOC states 80 and below 80 90 90 100 and 100 or higher This final state largely to be avoided simply provides a high terminal voltage so that the control algorithm is able to prevent overcharging Shown here are the resulting equations for the 90 100 SOC range within which the SuPER 52 team would like to operate the battery most of the time As in 19 multipliers are added to the R and SOC equations and the V open circuit voltage equation is slightly adjusted Discharge Mode V 1 95 18 P n R 19 et Hiss o B 14 SOC Charge Mode V 2 148 B n ee Ge ean 1 06 8B SOC SOC SOC SOC k V I SD SOC SOC t 0 625 m See the C code in Appendix D for the equations pertaining to the remainder of the states Den Herder s implementation includes a for loop with which the model integrates over time to calculat
61. gure 2 16 Phase Block Diari iodo aasscecersaueten uegncnstecensaanteaiauas vant 21 Figure 3 1 SuPER Software Flow DiagraM ooncocionionnnonnoncocnonannoncnncnnconononnonncnnnnonoco 35 Figure 3 2 SuPER Status and Control Interface Diagram ooooonconncnnoninccconnnoncnononancnnannnons 36 Figure 3 3 BP150SX I V and Power Curves 28 ooooooncccioconoccciccccoccconccconoconaconncconncnoo 37 Figure 3 4 USB Serial Cable with PL2303 Chip ce cecescssecseeneeeeeeeeceseceaeceeeaeeneees 38 Figure 4 1 SuPER Simulink MO 41 Figure 4 2 Simulink Model Map cicicioioinicn iaa dni 42 Figure 4 3 Simple Buck Converter ista da 43 Figure 4 4 a PV Array Voltage Response for Varying Insolation Levels 68W Load 44 Figure 4 4 b Example of Corresponding I V Curves oooconocccocccconcconncconoconnconnnonanccnnnonns 44 Figure 45 PV Array Mod A din 47 Figure 4 6 PV S function Block aii 47 Figure 4 7 Control S function Block dai la EN td RS 48 Figure 4 8 Switch Control S function Block ooninonnoninononnnonnninnoronanocnnnnonnononnncono nos 48 Figure 4 9 Deka VRLA Battery Self discharge Chart 15 ooooonconiccnncnincnoccnococancoccnonons 50 Figure 4 10 Current vs Capacity for AGM and Gel Batteries 000 0 cece eeeeeeteeneeereeeeees 51 Figure 4 11 Battery S function Bl d 51 Figure 4 12 Cooler S function Bl e O aii 54 Figure 4 13 Laptop S function Block sccjicicissesscissaaccnsisacsnccasanacastansneessus
62. he Figure 5 3 plot shows the result of four simulations each representing some number of hours after sunset for which the LED lights are powered Insolation is shown for perspective Note that time 0 in this case does not represent 6 00 AM but represents one hour before sunset instead r 1200 1 04 1 02 oN T ba four hours 1 i 780 E iria O gus t 600 32 3 E F x ies 5 two hours 5 030 i 4 one hour E 0 94 3 200 8 E 0 92 1 o 0 9 200 0 1 1500 time min Figure 5 3 Nighttime LED Operation Simulation As the lights are powered for a longer period of time the next morning must begin with less available battery charge and more time is required for the battery to reach full capacity The slight dip in the early morning SOC shows an hour or so passes before the sun alone provides enough energy to power the laptop Also the overcharging protection code dependent upon the battery voltage is the cause of the slight fall in SOC at the end of the day 68 As the heaviest load the motor will only be operated in daylight and for a short time For the sake of argument let us run the motor for one continuous hour per day The Simulink model can help determine when the most favorable hour for motor operation falls during the day It may be tempting to assume that running the motor at peak insolation approximate
63. he loads which SuPER will power and specifics about the team s approach to battery management which becomes a very sensitive issue Some of the requirements for the next generation of the system are presented which will provide an understanding of what the SuPER team hoped to accomplish by the end of March 2007 Chapter three provides the details for the software of the prototype s status and control system the real brains of SuPER The system consists of a series of sensors which feed data into a laptop computer for computations and convey corresponding output signals to control various components Chapter four represents the bulk of the work done by the author which was the creation of a MATLAB Simulink model of the entire SuPER system Each part of the model is presented and design decisions defended Chapter five describes through examples how the simulation is used to estimate the optimal control strategies to be put into use with the prototype s control system A comparison of measurements of the prototype in action and simulation results is made for multiple load scenarios The final chapter gives closure to things learned and conclusions determined from the experiences of this thesis work Recommendations are made for the future as part of a review on problems that have now been neatly defined thanks to the progress made on SuPER in the last six months Appendices at the very end of the document are a repository for information
64. he state derivative and h is the step size An implicit solver computes the state at the next time step as an implicit function of the state and the state derivative at the next time step e g X n 1 X n h DX n 1 0 This type of solver requires more computation per step than an explicit solver but is also more accurate for a given step size The following table lists the available solvers and the integration techniques they use Solver Class Integration Technique odel Explicit Euler s Method ode2 Explicit Heun s Method ode3 Explicit Bogacki Shampine Formula ode4 Explicit Fourth Order Runge Kutta RK4 Formula ode5 Explicit Dormand Prince Formula ode 14x Implicit Newton s Method and Extrapolation 97 The integration techniques used by the fixed step continuous solvers trade accuracy for computational effort The table lists the solvers in order of the computational complexity of the integration methods they use from least complex ode1 to most complex ode5 Choosing a Fixed Step Continuous Solver Any of the fixed step continuous solvers in Simulink can simulate a model to any desired level of accuracy given enough time and a small enough step size Unfortunately in general it is not possible or at least not practical to decide a priori which solver and step size combination will yield acceptable results for a model s continuous states in the shortest time Determining the best solver
65. hesis is a matter of construction detail and scale In the conclusion of his report Den Herder makes some observations and recommendations on improving the model all of which are addressed in the model presented here Figure 4 1 is a view of the entire SuPER Simulink Model 40 49081 En Paw nes aiez of auno area MSS puyo 292 eaten doder o uayosu enbe umeuo aun adus pinoy sawn aptues oll au adus yoa Sauer Aq TS nee L a ut ares fos hoder pora P A Pa al i i Bas Ey a gt de R sm ro dl g 4 qa E a i J Wd WW mue po PERE E i E pas T J A e a m a mw EY aN o e IA o o ERES z we ule Pa f Bl a le ZE m lt 4 o e p18 r aj ajqnop aw ardues fums Figure 4 1 SuPER Simulink Model 41 Each component of the model will be presented in detail For orientation assistance a generalized map of the key components and connections in the mode
66. hich contains air chilled to the same temperature Equalization with ambient temperature which again is fairly constant has been calculated to be approximately 008 F per minute The ratio of warming rate to cooling rate is 16 Thus to maintain an average temperature power theoretically need be delivered to the cooler for only 8 3 minutes of every hour This characterization is fine for very gradually changing external temperatures but proves all but useless for rapidly changing temperatures Figure 2 9 shows the results of a study done in which the cooler experiences three 60 minute power cycles each nine minutes on and 51 minutes off 17 100 90 gia ee 70 ee eb ae diffarbiert water E ml ambient cooler air f ae cooler water 30 P a OR 20 R 10 g T T T 7 o 50 100 150 200 time min Figure 2 9 Cooler 60 minute Cycle Temperature Study The cooler on times can be identified by the corresponding drops in interior air temperature One problem with this cycle period is the time taken for the interior air temperature to drop low enough to begin to cool the water a sort of setup time It is likely more efficient to increase the cycle period so that setup times are a smaller ratio to total cooling time A second study with measurements plotted in Figure 2 10 extends the cycle period to 725 minutes The cooling per
67. hree essential locations in the system the PV array the DC DC converter output and the battery In addition the voltage and current are monitored at each load Newly added in recent months is a pyranometer which outputs a voltage level corresponding to the level of insolation Therefore the total number of status system inputs M is defined as M 10 2 L where L is the number of loads The current sensors used are the ZAP25 and AMP50 models manufactured by Amploc Temperature data is provided by National Semiconductor s LM50 sensor 12 Vasquez s efforts included sensor characterization and calibration the construction of sensor circuit boards and work on the status system data reading code The sensor circuit boards are for converting the current and temperature sensor output voltages to proper levels for the A D inputs Also the subsystem voltage levels are stepped down to required levels for the USB 6009 devices The pyranometer is manufactured by Apogee Instruments Inc and measures the insolation which is the radiation between wavelengths of 300 and 1100nm incident to the Earth s surface 9 The level of insolation outside the earth s atmosphere has been measured at 1370 watts per square meter Wm 10 The level incident to the earth s surface is less due to atmospheric attenuation and other factors and the maximum terrestrial insolation observed by SuPER team members is just under 1100Wm There is a reduction in i
68. iod occurs in the first 100 minutes 80 70 50 dff ambient water ambient c A coder air 30 Cooler water 10 or E o o 100 200 300 400 500 600 700 time min Figure 2 10 Cooler 725 minute Cycle Temperature Study At time 538 minutes the cooler was brought indoors to provide a more constant external temperature for reference and comparison purposes The much more steady temperature difference rate of decline from that point onwards is clear It can also be observed that 18 the quickly increasing external temperature has a high impact on the rate of change of internal cooler air temperature when the cooler is unpowered Maintaining some average difference between ambient and water temperature is difficult under variable conditions If the cooler is to be run out of doors it may be necessary to power the cooler much more often than desired If that is accepted perhaps some power savings can still be achieved while measuring only the ambient temperature and compensating by adjusting run times 2 2 3 LED Lights One of the primary functions of SuPER will be to provide a few hours of night time lighting for the family home The necessary energy will be drawn from the battery It is expected that the control system will ensure that the day ends with the battery in a high SOC in anticipation of the energy requirements for lighting It is
69. irradiance double Iph G 0 Isc Define thermal potential Vt at temp TrK double Vt_TrK n k TrK q Define b Eg q n k double b Eg q n k Calculate reverse saturation current for given temperature double Ir_TrK Isc_TrK exp Voc_TrK Vt_TrK 1 double Ir Ir_TrK pow TaK TrK 3 n exp b 1 TaK 1 TrK Calculate series resistance per cell Rs 5 1mOhm double dVdI_Voc 1 0 Ns Take dV dI Voc from I V curve of datasheet double Xv Ir_TrK Vt_TrK exp Voc_TrK Vt_TrK double Rs dVdI_Voc 1 Xv Define thermal potential Vt at temp Ta double Vt_Ta n k Tak q 91 Ia Iph Ir exp Ve la Rs Vt_Ta 1 f la Iph Ia Ir exp Vc Ia Rs Vt Ta 1 0 Solve for la by Newton s method Ia2 Ial f lal f Tal la_new 0 Initialize Ia_new with zero Perform 5 iterations for j l j lt 5 j Ta_new Ia_new Iph Ia_new Ir exp Vc Ia_new Rs Vt_Ta 1 Ir Rs Vt_Ta exp Vc Ia_new Rs Vt_Ta Ipv 0 Ia_new ac D 2 Control S function Code ceontrol_plus_sre c function control block function to execute MPPT via pwm duty cycle of pv module and control load switches 1 bulk charge 2 float charge in Vpv Ipv Vb charge_mode out DC DCprev Pa 11 Written by Tyler Sheffield 12 10 06
70. is 93mH calculated using the equation for the desired maximum converter output current from 29 R AE IL V P A D is the duty cycle as a fraction of value one fsw is the switching frequency and R the load resistance The capacitor size C is determined by the desired ripple on the output voltage AV and found for this model with the help of some experimentation to be 3uF This is the relationship from 29 D V Mia R C f sw In a typical buck converter configuration with resistive loads the potential produced by the front end source Vs is bucked to a desired average output V by altering the switching duty cycle accordingly The relationship is V DV 43 For this application the output voltage is anchored to values near 12V on a range of about 10 14V by the battery Duty cycle adjustments will instead reflect on the converter input voltage which is the voltage Vpy at the PV array terminals This voltage in conjunction with the temperature and available insolation determines the amount of current output by the array Figure 4 4a shows the voltage that is seen by the array as a function of the duty cycle in the Simulink model of SUPER The relationship is not linear This plot shows curves for three different insolation levels all at a constant load of 68 W a typical load scenario Alongside is an example of how the array I V curves may look for these different levels
71. l is offered in Figure 4 2 Red lines represent electrical connections while blue lines are purely inter block signal lines duty cycle control DC DC Vov battery Figure 4 2 Simulink Model Map load switch In Simulink there are three methods of expressing the functionality of operational subsystems or modules component blocks mathematical function building blocks or MATLAB code also C C etc Den Herder used the function block construction method for the battery control and converter subsystems To assist the Simulink model in better reflecting its real life counterpart the converter was recreated by Dr Harris using the component blocks available in the SimPowerSystems package In a nutshell this means building the converter out of capacitors and inductors etc rather than representing it with a collection of mathematical function blocks For modularity reproducibility and optimization purposes the battery and control modules were remade as S function blocks of code 4 1 1 Design Approach Figure 4 3 shows the typical configuration of a simple buck converter 42 5 a ia Vo Figure 4 3 Simple Buck Converter The DC DC converter MOSFET switch is driven by a 500 kHz PWM signal a rate defined by the capabilities of the PIC 18F4320 The energy storage components an inductor and capacitor are the key converter parameters chosen based on the desired response of the converter The inductor value L
72. lt to model accurately A decision should be made in the near future on how much more time and money should be invested into the current battery and model As for incorporation of the battery temperature measurement it appears that thermocouple technology is the best bet for reading the temperature on the battery itself 84 1 2 3 4 5 6 7 8 9 10 11 12 Bibliography Harris James G White Paper for Sustainable Power for Electrical Resources SuPER July 15 2005 lt http www ece calpoly edu jharris research super_project white_paper_susper p df gt Tal Eran SuPER System Prototype Design and Implementation Master s thesis California Polytechnic State University 2006 Trends in Atmospheric Carbon Dioxide Chart National Oceanic and Atmospheric Administration 2007 lt http www esrl noaa gov gmd ccgg trends gt Venture Capitalists Embrace Solar Energy MSNBC 28 December 2005 lt http www msnbc msn com id 10625903 gt Mills Evan The Specter of Fuel Based Lighting Science 27 May 2005 1263 1264 Sharaf A M and A R N M Raez Ul Haque A Low Cost Stand Alone Photovoltaic Scheme for Motorized Hybrid Loads IEEE Proceedings of the 36 Southeastern Symposium on System Theory 2004 Chiang S J K T Chang and C Y Yen Residential Photovoltaic Energy Storage System IEEE Transactions On Industrial Electronics Vol 45 No
73. lties with this approach involve proper marriage of this signal to the MOSFET switch Casanova and Shein are using an entirely different method 25 SuPER has been provided with dual 75W buck converters as a donation from Linear Technology These converters have a built in PWM signal generation chip which uses a resistive feedback line to maintain a constant 12V output Some modifications were necessary to apply the device to SuPER as the output cannot be constant due to the battery and loads It was theorized that using different values of resistors on the feedback line would alter the response of the PWM generator chip and could be used to adjust the output voltage of the converter Casanova and Shein proved this true and a Maxim digital potentiometer MAX5529 controlled by a two wire serial interface is used to provide the changing resistance Its 64 tap configuration will allow a 1 56 duty cycle resolution Control of the potentiometer will be via the digital output ports on the USB 6009 device Code written to communicate with the potentiometer potcomm c has been tested successfully Perhaps the most useful of the proposed power sinks for SuPER the LED lighting remained an untouched matter through the summer of 2006 LED lights are a continually evolving and also pricey technology nevertheless the Phase system provisions their inclusion The lights are the final of the five proposed loads the prototype will service in these experimental
74. ly 13 30 to 14 30 is the best option However it may prove wiser to operate the motor in the morning hours and take advantage of the afternoon sun to recharge the battery Figure 5 4 is a plot displaying the battery SOC over the course of the day for different hours of motor operation peak straddle o S off peak one O off peak two 8 off peak three T k off peak four E off peak five A Jaan G time min Figure 5 4 Motor Operation Simulation We wish to drain the battery as little as possible and still be able to recharge it fully before the day is through According to these simulation results the hypothesis may prove correct Operating the motor at around one to two hours before peak insolation should only deplete the battery to a little below 92 and enough sunlight time will remain for a full recharge The inconsistency in the recharging curves of these various 69 scenarios can be attributed to the difficulty in modeling the battery while in the current limiting float charge stage 5 2 Result Validation The next step towards proving the value of the simulation is to compare actual prototype system measurements to simulated versions of equivalent operating conditions The preliminary exploratory simulations have assisted in defining what kinds of tests should be run
75. mall If you changed the maximum step size try running the simulation again with the default value auto Setting the relative tolerance too low can slow down the simulation The default relative tolerance 0 1 accuracy is usually sufficient For models with states that go to zero if the absolute tolerance parameter is too small the simulation can take too many steps around the near zero state values The problem might be stiff but you are using a nonstiff solver Try using ode15s Mixing sample times that are not multiples of each other causes the solver to take small enough steps to ensure sample time hits for all sample times Smaller steps lead to longer simulation times Be sure to eliminate algebraic loops if possible The solutions to algebraic loops are iteratively computed at every time step Therefore they severely degrade performance 101 Appendix I SuPER Prototype Operation ON Seo Ensure that all breakers are open Insert the hub cables into the laptop USB ports followed by the NI DAQ device cables Then insert the PIC cable into the open laptop port The mouse cable should be inserted into the hub Power on the laptop at this point running on its internal battery and at the GRUB window choose the latest version of Red Hat Login using root super1 Open a shell and change directories cd to home super1 pvpro src Close PV converter and battery circuits by flipping the breakers marked PV B
76. n USB 6009 Devl and data recorded and stored by the computer software In software the amplifier gain A will be removed by division and the resulting raw value in millivolts will be multiplied by 5000 to give insolation G in Wm g _ multiplier W m V V 5 1000 V 7 A 552 All data produced by the status system is collected by the control system at a rate fis defined as the status system sampling frequency Its inverse is Tss which is currently set at two seconds There are N control system outputs where N is defined as N 3 L where L is the number of loads One of the outputs is the value of the duty cycle for the buck converter PWM signal This value is usually spoken of as a percentage with a minimum 14 of 0 and a max of 100 In the PC software it is stored and manipulated as a floating point number with values between 0 and 1 When transmitted serially to the PWM producing microcontroller the value is first represented as an 8 bit unsigned binary number The microcontroller code provides the proper mapping of the unsigned number to the desired duty cycle of the PWM output The remainder of the outputs produce binary on off values These control the MOSFET switches that dictate the flow of current in the system There is one switch each for the PV array and DC DC converter and one for each load circuit 2 2 SUPER Load Characterization 2 2 1 Television The television is the simplest of all loads considered The unit
77. ncident insolation as the angle between the normal to the sun s rays and the line of propagation of rays to the point of measurement increases Thus equatorial regions receive greater insolation than other regions of the planet in general The amount of reduction corresponds directly to the angle and this effect is expressed mathematically as Lambert s cosine law 9 for this reason Apogee instructs that the pyranometer must be mounted parallel to the ground The pyranometer is calibrated to output 1mV per 5Wm or a maximum of 220mV at the full insolation level of 1100Wm The value of 5Wm in this ratio is determined by fabrication methods and materials and is inscribed upon the device by the manufacturer The manufacturer reports a temperature sensitivity of about 1 per degree C for which we do not compensate at this time A circuit for amplifying the pyranometer output was constructed on a breadboard attached to the inner wall of the switch box an effort supported by senior Slavic 13 Orzhakovsky The circuit is diagrammed in Figure 2 6 We use an LM324 operational amplifier in a voltage reference configuration powered with an LM340 voltage regulator which steps down the 12V system bus voltage to 5V The measured resistor values are 9 87kQ and 2 18kQ for R and Ro respectively This results in a gain A of 5 52 Figure 2 6 Pyranometer Data Circuit The output of the pyranometer amplifier is fed into analog input 7 o
78. nd accompanying values The microcontroller sends no data to the PC but does respond to successfully received commands and values by returning an exclamation point character UART serial communication is byte oriented and for ease of implementation all commands and values are eight bits in length or less An explanation of the communication protocol can be found in Appendix C The battery model code from the Simulink model has been ported to the laptop in the form of a function called batt_voltage in contAcquireNChan c It currently monitors battery current flow to estimate the actual battery SOC in real time Output is written with frequency fss to Super_Output csv 38 The prototype control software is largely incomplete as some of the hardware goals for the end of March 2007 were not attained The only form of control currently implemented in the prototype software is the MPPT algorithm even so the generated output is actually of no practical use without a DC DC converter The prime resource for developing and testing control algorithms then is currently the Simulink simulation 39 Chapter 4 MATLAB Simulink Model 4 1 Model Overview The new Simulink system model builds upon the foundation established by Tyson Den Herder in his senior project 20 but attempts to reach far beyond its limits and uses a different development approach The primary difference between Den Herder s efforts and what is to be accomplished in this t
79. nificance Type k battery efficiency constant SD h self discharge rate constant Ns number of series 2V cells constant SOC initial SOC percentage constant SOC Wh battery capacity variable SOC Wh estimated remaining energy variable B SOC SOCm variable I A battery current variable The model uses these parameters to predict the battery internal resistance R and terminal voltage Vpat at time t Den Herder uses a 12V 66Ah 20 hours 792 Wh capacity battery SuPER s Deka 8G31 model is rated at 97 6Ah 20 hours 1171 Wh so the model must be updated accordingly The charge discharge efficiency value k is not made available by the battery manufacturer so following one of the examples given in 19 and echoing Den Herder s choice a conservative value of 0 8 will be applied This coefficient is a multiplier of the battery current in the model SOC equation so adjusting it will impact the rate of change of the SOC in both charge and discharge states for charge associated with the current flow This parameter was not used in making adjustments to the battery model because of the need to account for differences in charge and discharge behavior It is also necessary to adjust the self discharge rate which is provided by the manufacturer According to the specifications Figure 4 9 the battery will linearly lose 49 50 capacity over a 16 month period assuming it is sitting at typical room temperature 20
80. nnocinnss Figure 5 9 March 19 2007 Motor Simulation c csccsscssssessssesssesssessssesssesseesseesssessseen Figure 5 10 March 19 2007 SOC Estimates c ccssscsssesssesseessessseessesseessesseesseesseesses Figure 5 11 March 29 2007 Insolation and Temperature scscccsssesssesseeseeseeseessee Figure 5 12 March 29 2007 Motor Operation Measurement c0scsssesssesseeseeseee Figure 5 13 March 29 2007 Motor Simulation c sccsscesseessessseessesseessesseesseesseesseesees Figure 5 14 Five Load Two Day Scenario One a Load Schedule b SOC ESTA e Et o eee eee Figure 5 15 Five Load Two Day Scenario Two a Load Schedule b SOC ESTA Rao Figure 5 16 Four Load Two Day Scenario Three a Load Schedule b SOC A a ay gatdee a igs emo AE E A Et Figure 5 17 Simple Resistive Loss Model ini da Figure 5 18 System Power Levels 70W Load on CKT 3 0 eee eeeeseeeeceseceeeeneeeneeeeeees Figure 5 19 PV Power and Converter Efficiency ooooonoonnncnonnnonnnannnncnoronnnnoranoncrnnconocos viii List of Tables Table 2 1 Deka Battery Charge Voltage Guide 15 ooooooonccnnncninccnoccniccconanconnaconocnnos 25 Table 2 2 DC DC Converter Specifications 2 oonnnccnnncninnoninccincnconaconncnonaconncconocnnos 2T Table 2 3 2006 2007 SuPER Project Student Contributions ooooonconoccnncnnnncnconncconocnnos 30 Table 3 1 Known USB 6009 Errors ccceccces
81. o bulk and float charge stages respectively The stage is adjusted according to the battery voltage V Port Identity DC duty cycle initialization Ipv PV current A Vpv PV voltage V Vb battery voltage V cm charge mode initialization 1 2 DCout new duty cycle DCprev old duty cycle for debug Ppv PV power W for debug cm_out old charge mode 1 2 count mode restriction 0 1 for debug Figure 4 7 Control S function Block In the switch control S function block load operation decisions are made For this early version of the control system a table is created that holds the on and off times for each of the five loads The code then uses the system time to flip the load enabling switches on and off There is a scenario selection input that lets the model user identify which load time table to use This block is shown in Figure 4 8 Port Identity scenario load scenario identifier 0 13 gt stime system time sim minutes switches 0 4 load control output 0 1 Figure 4 8 Switch Control S function Block 48 Of special relevance to SuPER is the battery model in use Den Herder s simulation uses the PSpice model from 19 adapted to Simulink Table 4 1 identifies the parameters associated with this model Table 4 1 Battery Model Parameters Parameter Units Sig
82. oblem with severe downsampling and eliminated all signal probing at less important locations Simulink s downsampling blocks available in the Signal Processing toolkit allow the user to specify the downsampling ratio and offset Later a more elegant solution was discovered in the scope blocks themselves The scopes can be instructed to perform decimation on their inputs see Appendix F We must see at least one sample for each event that alters the steady state of the system since we are not interested in tracking the details of the transient system response Thus the maximum amount of decimation is determined by the block with the least sample time L according to the following Dez shortest block period _ L systemsample time T Of course decimating at the maximum and running a simulation for less than T in duration will result in zero data points Steps for accessing detailed simulation characteristics via Simulink s coverage reporting capability are found in Appendix G 64 At this stage all major barriers to running a successful simulation have been overcome Though there are several minor tweaks and improvements that can be made simulation results have provided encouraging validation for the usefulness of this model 65 Chapter 5 Observations and Model Authentication 5 1 Exploratory Simulations We wish to operate all loads as much as possible however SuPER s ability to do so is dependent upon the po
83. ocnnoccnoncconncconaconn nono ncnnnccnocnnos 66 5 1 ESPIAS AO E A io 66 5 2 Result Vadis 70 5 3 Multia oad Scenarios e 75 CONE sacra E T E T E A E EE AST 78 Chapt r 6 Concl sioN ienero anis e ad ed 81 6 PICNIC V CISTI G A E A A E A a E E S 81 6 2 Reflection on System Sensitevities id dos 82 6 3 IRECOMIMENC A ONS terin nnen AR E O E E TE a nies 83 A e A pee 85 Appendix A NI DAQmxBase 2 1 API Function LisSt oooooncnncnnncnncnnnccncnconannncnanonccnnnons 88 Appendix B Status Data Extraction Macro for ExC l ooooonnonnonicnnncinccnocconnconccononanonncnnnons 89 Appendix C PIC Serial Communication Protocol ooooccnnccnnccnocccononononcnnncconocono nono nonnnonns 90 Appendix D C MEX S function Code wcities n 91 D 1 PV Array S function osas 91 D2 Control S f ncton Cod rs esinaine e e a aeeie 92 D 3 Switch Control S function Code iii asi 93 D4 B ttery S f nction CUA asrine e a E riadas 94 DS Laptop S f nction Codes ii a Sc 95 D 6 Cooler S function Code acia 96 Appendix E Choosing a Fixed Step Solver irradia inc dic 97 Appendix F Managing Scope Data bd cays 99 Appendix G Specifying Coverage Report Settings ooooonoccionnnnccnoncconaconnconanconcccnnocnnos 100 Appendix H Improving Simulation Performance and ACCULACY coooococcocccocconcconanononnnonns 101 Appendix I SuPER Prototype Operation cccecccccsseceseceeeceeseecseecesecseeeeeseecsseceeeees 102 Appendix Je Ele README id eS Asa 103 vi List of Figures
84. otor Model Simulation with Parallel 58F Capacitor time in s 4 2 Principles of Timing and Sampling One of the key issues confronting the creation of a system simulation is the handling of the various rates of system elements such as the MOSFET 500 kHz switching frequency the rate of environmental data sampling the control system operation rate and the battery and load data update rate We would have preferred to run the simulation in continuous mode with a variable step solver for the sake of accuracy However such simulations have proven to be far too computationally intensive and time consuming to be a realistic option A discretized simulation is necessary but fraught with its own perils Choosing a solver can be a frustrating issue Simulink has a variety of available continuous and discrete time solvers and it is not always clear which one will serve the model s purpose best Appendix E contains information on choosing solvers distilled from MATLAB s user guide 31 For SuPER it was realized that no continuous states were necessary in the model and the fixed step discrete solver was chosen however the model has matured enough now that many of the fixed step solvers appear to be viable 58 Variable step solvers are not an option as they do not tolerate the presence of the running mean blocks and choke on mixed sample time errors There is a delicate tradeoff between the system sample time and the resolution for the dut
85. ower generation and distribution it was clear that the call for a digital control system could use some computer engineering expertise I began meeting with the development team in January 2006 which is about the time construction of the project prototype began My contributions to the effort for the first six months of my association with the project consisted largely of support for Eran Tal working on his thesis 2 Readers new to SuPER ought to become familiar with Tal s work as he led the team in building the first stage of the prototype and provided a foundation for all that has been accomplished since During this period I provided some software expertise to the team in doing some C and assembly language programming as well as managing the Linux development environment on the project computers Since Tal graduated in summer 2006 and I took over project leadership SuPER has been a whirlwind learning experience for me My educational and professional engineering experiences have largely fallen under the programmable logic embedded systems and signal processing disciplines I have never been a power and control systems engineer or a proficient analog circuit designer yet while working on SuPER I have found a need to be a little bit of each of these in order to reach both personal and team objectives That is perhaps the most rewarding part of the entire experience The knowledge gained working on this project has added significant breadth
86. p by adding a fixed step size to the time of the current time The accuracy and length of time of the resulting simulation depends on the size of the steps taken by the simulation the smaller the step size the more accurate the results but the longer the simulation takes If you allow Simulink to choose the step size Simulink sets the step size to the fundamental sample time of the model if the model has discrete states This choice assures that the simulation will hit every simulation time required to update the model s discrete states at the model s specified sample times The fixed step discrete solver has a fundamental limitation It cannot be used to simulate models that have continuous states If you attempt to use the fixed step discrete solver to update or simulate a model that has continuous states Simulink displays an error message Thus updating or simulating a model is a quick way to determine whether it has continuous states The continuous solvers employ numerical integration to compute the values of a model s continuous states at the current step from the values at the previous step and the values of the state derivatives Simulink provides two distinct types of fixed step continuous solvers explicit and implicit solvers Explicit solvers compute the value of a state at the next time step as an explicit function of the current value of the state and the state derivative e g X n 1 X n h DX n where X is the state DX is t
87. period on the order of a few minutes or less can cut battery energy costs down significantly A 58F ultracapacitor manufactured by Maxwell Technologies was purchased and will be introduced into the system The motor load is circuit 6 on the prototype 23 2 3 Battery Management The SuPER prototype battery is a 12V valve regulated lead acid VRLA unit manufactured by East Penn or Deka rated at 98Ah for 20 hour discharge The SOC of a lead acid battery is a percentage representing the ratio of charge remaining to total battery charge Its inverse is the depth of discharge DOD Determining the SOC for VRLA batteries while connected to a load has always been a difficult problem The simplest most typical way to make this determination in practice is to measure the open circuit voltage Voc of the battery due to a nearly linear relationship between Voc and SOC for lead acid batteries 15 This method provides a reasonably accurate assessment however it is unrealistic for many systems because a true Voc can only be attained after all current flow in and out of the battery has been suspended for 24 hours 15 This is not an option for the SuPER project The general approach to this problem is to use frequent measurement techniques to estimate the SOC in software Methods to this end are proposed by Vairamohan in 17 Duryea Islam and Lawrence in 18 and Castaner and Silvestre in 19 The SuPER team has chosen to use the model in 19
88. r an introduction to this macro After each data set is observed and averages calculated a call is made to pnopal c for running the control algorithm pnopal c contains the MPPT algorithm and sends the new duty cycle value to the PIC by calling commpic c commpic c is the code that provides serial communication from the PC to the UART on the PIC Figure 3 1 is a software flow diagram that summarizes all the simultaneous processes in execution when the status and control software are running 34 USB 6009 RTOS forked main pnopal c commpic c Microchip PIC ate Sample inputs ii TPN from storage el et command aN assigned Paai o digital values to P amp O Define create and start sampling and digital write time reached pr lt a Send aN read Ag D VSN task write Tra ump al accumulated data to hard NY Kill a cy MPPT Y fas s new duty cycle value to comm ento modem p d e gt cycle value for Se wa j ransmit value serially 1d ower on pwm initialization set swit ching requenc UART command rxed t duty cycle to Figure 3 1 SuPER Software Flow Diagram 35 3 3 Control The diagram of Figure 3 2 details the locations of all Phase 0 1 control inputs and outputs Vb Ib Tb
89. r input due to the manner in which the PV array is modeled for simulation purposes Voltages and currents plotted with Simulink scopes oscillate at the switching frequency The output voltage ripple is intended to be minimized by careful attention to the capacitor chosen for the converter to smooth the output to a greater extent and allow the developer to see some semblance of average DC values Simulink offers two options neither of which are ideal for this use One option is the weighted moving average block Weights need to be assigned to each sample and the number of weights determines the window size This makes it impractical for windows of thousands of samples and has proven difficult to use in practice The chosen method is a reset enabled running mean block whose reset period is a confirmed hazard in simulation The optimal period seems to be twice the fastest S function block sample period The running mean blocks must be reset on the falling edge of the reset signal so that any downsampling does not catch the block output early in the new mean processing as the oscillations will result in unsettled sampled values 45 4 1 2 Function Blocks Desiring a modular simulation model we made prolific use of MATLAB s S functions These are created as blocks in the Simulink model editor but completely defined by MATLAB code in associated m files There are a variety of S function types but the original system design was done with
90. rial interface to a digital potentiometer 30 Chapter 3 Prototype Software 3 1 Interface The status and control system for the Phase 1 prototype is all managed on a Dell Inspiron B120 laptop computer This machine is equipped with an Intel Celeron M 1 4 Ghz processor and 256 MB of DDR SDRAM With a 40 GB hard drive it is more than sufficient for SuPER s computing power and data storage needs The laptop executes all data acquisition and control code over a Red Hat Enterprise Linux WS 3 operating system A few factors figure into the decision to use a Linux platform First one of the goals of the SuPER team is that all software for this project be developed as open source and protected under a general public license GPL This will ensure that the work will be available for modifications and expansion as well as learning purposes for any who may want to take advantage Second Linux facilitates C development in general better than other platforms and for this project the ease of access to system level kernel function calls is of paramount importance Thirdly the project team at the time felt most comfortable developing in that environment due to significant previous experience with Linux NI provides a well documented C application programming interface API to accompany their multifunction data acquisition DAQ devices 26 The name of the package is NI DAQmxBase 2 1 This API consists of C functions that provide direct access
91. s The SuPER team attempts to mimic this capability and will also run the control algorithm at the maximum rate It is anticipated that the host machine will have adequate time to run the few necessary floating point multiplications and divisions between samples and that computational time overruns will not be an issue The DC DC converter transient response discussed in more detail presently will not be an issue at this rate The prototype currently makes use of a PIC 18F4320 microcontroller which can provide a 500 kHz PWM output This is an upgrade from the 50 kHz signal provided by the original Phase 0 hardware a PIC 16F877A The PIC code is written in assembly language and compiled with the MPLAB development kit provided free of charge by 37 Microchip Programming is achieved with the K128 USB 40 pin programmer from DIY Electronic Kits http www kitsrus com pic html The laptop is not equipped with a serial port so the connection to the PIC is accomplished via a USB to serial conversion cable The cable manufacturer is unknown but the conversion chip is a product of Prolific Technology Inc the model number is PL2303 Use of this cable in Linux requires driver installation and configuration Figure 3 4 USB Serial Cable with PL2303 Chip It was necessary to develop a simple communication protocol for all serial transmissions between the PC and the PIC The communication is largely one way as the PC issues all commands a
92. sing the Windows XP battery meter and displayed against time in Figure 2 12 oa885883888 SOC 4 ie 20 40 60 80 100 120 time min Figure 2 12 Laptop Battery SOC Under AC Power When powered by the SuPER cart without its internal battery the laptop draws approximately 2 5A as considered from a system perspective and hence concomitant to a potential of 12V With the laptop battery inserted the behavior seems to change relative 20 to the SOC of the battery A depleted battery will perforce need to be charged so the laptop will draw enough current to run the device and charge the battery in tandem The laptop circuit will in this case draw two extra amps for a total of 4 to 4 5A which is closer to the specified maximum power requirement With enough time passed to anticipate a fully charged internal battery it is observed that the laptop has again reverted to a 2 5A current draw Recorded data see Figure 2 13 shows that there is a gradual drop off in the current drawn by the laptop Using the data from Figure 2 12 we can predict the time at which the battery reaches approximately 80 of charge capacity 60 Brea 55 50 S 45 40 E 30 25 20 0 50 time rin 100 150 Figure 2 13 Observed Laptop Power Needs Under Solar Power This is
93. stages and senior Joey Zukowski was tasked with equipping the devices for SuPER at highest energy efficiency 28 An additional Phase goal is the development of a user independent control system which derives maximum use of each load while optimizing the life of the battery and preventing overcharging Essential to the development of an optimal control system is a thorough understanding of system behaviors under a variety of conditions It is therefore desirable to simulate the system and create a platform upon which control schemes can be developed assessed and adjusted as necessary This ambition became increasingly important to the project as it became clear that the integration of the DC DC converter would not be reached on schedule As mentioned previously there was a misstep in plans for handling the system s current requirements on the switch board in the Phase 0 system For a completely operational Phase 1 system the issue must be solved The team also determined to take advantage of these efforts to simultaneously increase the modularity of the system components specifically 1t would be valuable to physically separate PCBs of different purposes and current levels As summarized in Table 2 3 besides the work on the ultracapacitor seven parallel efforts were made from October 2006 through March 2007 to reach the Phase 1 plateau Some of the work by these Cal Poly seniors will require the inclusion of more digital control system o
94. stinguish between periods of differing levels of insolation and this is done primarily by monitoring the power produced by the array Despite the fact that for development purposes the insolation measurements are available to the control system it is not anticipated that an installed system will be accessorized with a pyranometer Thus control decisions will not actually be made based on measured solar insolation Of paramount importance to the project is extending the life of the battery so the two key factors in load operation will be battery SOC and power produced by the array Ppy which is directly affected by the actual insolation level Note that the prototype uses a laptop for all status and control operations so this will be taken into consideration for all simulations Thus it is presumed that the laptop will be drawing power during all daylight hours during which a load might be operating and any nighttime hours during which it is planned to run other loads particularly lights 67 We will operate under the assumption that the laptop s on off state is controlled intelligently by an external entity such as a human user There are many questions the simulation can answer for us which we can then verify through prototype operation For example it will be important to know how long the LED lights can be run in the evening given the stipulation that the battery should be able to be recharged to full capacity the following day T
95. table 5 INA table 6 INA table 7 INA table 2 0 60 table 2 120 table 2 2 INA table 2 3 INA table 2 4 INA table 2 5 INA table 2 6 INA table 2 7 INA table 3 0 0 table 3 1 120 table 3 2 720 table 3 3 INA table 3 4 INA table 3 5 INA table 3 6 INA table 3 7 INA table 4 0 INA table 4 INA table 4 2 INA table 4 3 INA table 4 4 INA tablel 4 5 INA table 4 6 INA table 4 7 INA if int scenario 0 1 93 table 0 0 INA table 0 INA table 0 2 INA table 0 3 INA table 0 4 INA table 0 5 INA table 0 6 INA table 0 7 INA table 0 0 table 1 100 table 1 2 INA table 1 3 INA table 4 INA table 5 INA table 1 6 INA table 1 7 INA table 2 0 INA table 2 INA table 2 2 INA table 2 3 INA table 2 4 INA table 2 5 INA table 2 6 INA table 2 7 INA table 3 0 0 table 3 1 INA table 3 2 INA table 3 3 INA table 3 4 INA table 3 5 INA table 3 6 INA table 3 7 INA table 4 0 INA table 4 INA table 4 2 INA table 4 3 INA table 4 4 INA tablel 4 5 INA table 4 6 INA table 4 7 INA for i 0 i lt 5 it i is the load index for 4 0 4 lt 8 j j is the time value index if fabs table i 3 stimel000 lt 1 double type adjustment switches i switches il flip switch break
96. ted in MATLAB code for simulation Load Figure 4 5 PV Array Model 28 The insolation temperature and array voltage are fed to the PV array S function which simply provides a wrapper for O1 s BP150SX solar panel m file For the new C code blocks the solar panel code was translated to C The PV block outputs the current produced by the array which is built electrically as a controlled current source driven by the S function output Figure 4 6 shows the PV S function block Port Identity G insolation Wm TaC Temperature C Vpv PV voltage V Ipv PV current A Figure 4 6 PV S function Block Note that although the array is rated by the manufacturer at 150W peak in practice the SuPER team has observed a maximum output of only 122W at peak insolation An adjusting coefficient has been added to the array code to reflect this The control algorithm with P amp O code for the MPPT is contained in an S function block titled Control Figure 4 7 The Control block also requires initialization and 47 knowledge of a couple of values as MPPT operation is dependent on the system s behavior under previous outputs Thus the current duty cycle parameter DC which is not to be confused with direct current voltage current and charge mode cm are assigned initial conditions and fed back through Memory blocks The charge mode parameter has two possible values 1 or 2 which correspond t
97. tem 9 Q 0 92 0 9 0 88 328 378 428 478 528 578 time min Figure 5 10 March 19 2007 SOC Estimates In reality the measured battery Voc taken 24 hours after operation was suspended indicated a fully charged battery In consistently being conservative while tweaking the battery model it is possible that the battery s capabilities have been underestimated good news in terms of the viability of SUPER However it was also observed that the simulation tends to predict higher currents and voltages than the prototype battery actually experiences Perfecting the model will require time and careful attention to detail Another motor test was performed on the 29 of March conditions for which are found in Figure 5 11 1200 40 1000 A aa lt 30 A 800 25 G 600 l 20 3 E o S 400 ae E 10 2 200 5 0 i 0 273 323 373 423 473 523 573 time min Figure 5 11 March 29 2007 Insolation and Temperature 73 This time the motor is run from 12 30 to 13 30 which is the peak insolation period for this time of year in San Luis Obispo and the system measurements are shown in Figure 5 12 S a 5 g gMh eerie _ l p E 5213 323 373 423 473 523 573 623 E z 10 2 Es per o 20 time min Figure 5 12 March 29 2007 Motor Operation Measurem
98. that when that day arrives new PV cells and batteries can simply be inserted into an already proven digitally controlled distribution system To justify the cost of the new cells and batteries the balance of the SuPER infrastructure must be as economical and efficient as possible 82 Table 6 1 outlines the project costs to date Loads are part of the development cost but are not part of the 500 target for the end user system cost Table 6 1 SuPER Development Costs to Date Infrastructure Unit Cost Dell Inspiron B120 Laptop 450 Lind Electronics DC DC Converter 140 BP 150SX Solar Panel 750 12V Gel VRLA Battery 98 Ah 20h 150 NI USB 6009 DAQ Devices 420 Wiring breakers connectors etc 460 PCBs 400 Loads GPX Portable 5 television 15 Coleman 12V DC Refrigerator 90 LED Lights x4 70 Dayton DC motor 275 These costs total 2 074 for the SuPER infrastructure with nearly 3 000 in developmental expenditures up to this point One of the key cost cutting measures will be the replacement of the laptop and NI DAQ devices at the core of the status and control system Eventually we would like to see a low power FPGA take on all status and control duties This alone would reduce costs by nearly 1 000 As the system takes shape wiring and parts costs will be reduced significantly and PCB manufacturing processes will have the same effect Certainly the battery and PV array will be the most costly
99. that as soon as reasonable the original switchboard should be replaced by smaller modular boards each crafted to handle certain amounts of current Such modularity would have the added benefit of improving troubleshooting and repair turn around time As part of this process it will be necessary to learn about the design and manufacture of high current PCBs Kaha Sariashvili joined SuPER in January 2007 to design and test a board for the motor load and suggest new designs for the switch board 11 2 1 2 Status and Control Hardware A D conversion for data acquisition is accomplished by use of multiple National Instruments NI USB 6009 Multifunction Data Acquisition DAQ devices As indicated they interface to a PC host via USB All data that provides system status information to the PC for control comes in through these devices The network of sensors data acquisition devices and PC software that manages the devices and data is collectively known as the SuPER status system Figure 2 5 shows a simple block diagram of this system The hardware of this system in its current state is partly the work of Gustavo Vasquez as documented in his Spring 2006 senior project paper 8 PC wiLinux Operating Machine Data Acquisition Device NiDAQ Sensor Board emperature Current Figure 2 5 Status System Interface Block Diagram 8 There is a triple of key sensors voltage current and temperature at each of t
100. the model we were forced to clock the function by only allowing access to the mathematics on the edge s of a pulse signal The outputs are then only evaluated once per instance at a rate we specify For the C MEX S functions a function execution sample time can be defined This is accomplished by a setting on the Initialization tab on the S function dialog boxes The block sample mode is set to Discrete and the sample time directive defines the clock period The PV array S function cannot be clocked or sampled at a rate less than the discrete system sample time unlike other S functions because the array would be unable to respond properly to the system changes which occur at high rates due to the converter switching frequency Insolation and temperature data for a 24 hour day are stored in blocks of 1 440 samples supplying one sample per minute Since the MOSFET switches at 500 kHz the discretized simulation sample rate must be at or above the Nyquist rate of 1 MHz Running a simulation on such a scale yields a terrific number of data points unwieldy for 60 the PCs we are using We must therefore fool the system by decreasing the insolation and temperature sample times artificially For example we ll take one day s worth of data but tell the simulation that it is one second s worth instead As long as the transient response of the converter Figure 4 19 is not interfered with the simulation time can be greatly red
101. to and control over the devices With these functions the user can for example define and start analog input sampling tasks and set digital output values Appendix A taken from 27 outlines the API 31 The team encountered some trouble with Linux in regards to the integration and interface for the USB 6009 devices and the lessons learned are mentioned in passing here The Targus 4 port hub uses USB 2 0 drivers so it is essential that the latest version of the Linux kernel be installed on the host machine Version 2 4 21 37 is not equipped with the proper drivers and therefore version 2 4 21 47 must be installed Before halting execution of the interface software process all tasks assigned to the devices must be stopped and cleared Bypassing this step causes a glitch that will result in Linux losing the device identifiers restoring functionality requires a device hard reset disconnecting the devices from their USB power source host Unfortunately the example code that NI ships with the devices and upon which the SuPER code was built seems to disregard this peculiarity As the NI code is executed the user receives instructions indicating that the process may be terminated by using the ctrl c command This is the universal Unix process halt command This command does not allow the process to exit gracefully but ends its life by kernel override As a result the kernel somehow loses communication with the DAQ devices The problem w
102. uced Such a change will also affect real time values in hours used in the battery SOC and load current draw calculations so an adjusting time coefficient is included in those functions Figure 4 19 Discretized Converter Transient Response time in ms Here we are interested in the time scale the values shown on the x axis are milliseconds Thus the transient response is shown to be well below 50 us The short response allows the simulation time to be decreased significantly For the simulations in chapter five of this paper one minute in real time is equivalent to one millisecond in simulated time The discrete system sample time and switching frequency factor into the speed at which the simulation can be calculated The chosen values result in a simulation that takes approximately one second in real time for each millisecond in simulation time At this rate a simulation of 24 hours of data can be completed in about 24 minutes The governing factor for the control system update rate is the rate at which we wish to change the PWM signal duty cycle For the prototype we would like to change the duty cycle as fast as the system sample time Tss which will be the maximum rate In 61 Simulink since we want a real time minute to be as short in length as possible we must severely cut back the number of duty cycle adjustments per minute so as to maintain the 24 minute completion time for the simulation Table 4 2 holds the final values for
103. uffers and the NUM TO OUTPUT definition fixes the number of samples that are extracted NUM _ TO OUTPUT must be less than or equal to the integer bufferSize The extracted values are all averaged to formulate the display quantity The loop is executed at the system sample rate fss At the beginning of each loop cycle the buffers are checked to see if the write time has been reached as defined by TIME FACTOR in minutes The values must be periodically written to the hard drive so as not to be lost The written values consist of all samples extracted prior to averaging Thus the written files contain NUM_TO OUTPUT Tss samples per sensor for each second of run time and the total number of samples per sensor in the file is TIME FACTOR 60 NUM_TO OUTPUT Tos 33 Hard drive accesses are expensive operations and it is important not to delay the time sensitive loop commands so as to avoid the risk of device memory overflow Therefore main is forked so that a child process may take care of the file I O and the parent can return promptly to data reading Sensor data is written to the hard drive in comma separated value csv files The files are named with date and time included e g SuPER Wed Jan 10 10 44 30 2007 csv and stored in a brother folder to the source code entitled data To ease the manipulation of these large amounts of data an Excel macro has been created that consolidates the file into one minute samples See Appendix B fo
104. ut highly abnormal and great cause for alarm in this particular case In addition it was unexpected to see such a draw because some measure of protection had been expected from the circuit breakers in the breaker box However the team failed to account for the slow response of breakers to negotiating currents above breaker ratings Breakers do not in fact mimic fuses in functionality Their response instead obeys the tripping time curve seen in Figure 2 4 Thus for the 30A breaker in use on the hot battery line a 39A draw is just under 130 of rated current At that level it would have taken one minute for the breaker to trip It was never given a chance as the problem was discovered and the system shut down manually in a matter of a few seconds 10 1000 CRERAING CHARACTERSIICS AMBIENTTEMPERATURE 30 C 100 CURVE 9 ACIDO MENIUITES TRIPPING TIME SECONDS 0 07 0 00 8838 8 8 98 889 898 88 B eis Oy Mo g ON ee spe ane ep RATED CURRENT Figure 2 4 Circuit Breaker Industries CBI Breaker Response Curve One particular breaker on the DC motor load line was replaced by a fuse Largely motivated by this experience it was determined
105. utputs in the near future Zukowski will develop a DC DC converter to step down from the 12V bus voltage and deliver about 4 5W to four LEDs The PWM will also be transmitted by the PC through the PIC and managed by voltage and current monitoring code in order to achieve maximum efficiency with the LEDs The PIC can provide two PWM outputs if need be For immediate integration and testing a purchased static output buck converter will provide satisfactory output 29 Table 2 3 2006 2007 SuPER Project Student Contributions Project Student Contributors DC DC converter development Robert Casanova device modification Joe Shein DC DC converter development Thaddeus Guno computer controlled Koosh Shah Kunal Shah High current PCB development Shane Murphy thermocouples Juan Uribe Pyranometer integration Slavic Orzhakovsky High current PCB development Kaha Sariashvili Simulation and software control Tyler Sheffield LED lights subsystem integration Joey Zukowski Ultracapacitor integration Joseph Witts denotes independent study as opposed to senior project contributors Though his efforts are not associated with the Phase 1 objectives Joe Witts will add an ultracapacitor between the battery and the loads and will need control signals for three switches to manage the charging and discharging of the capacitor The main converter built by Casanova and Shein will require two digital outputs for a se
106. wer that can be harvested from the sun There are certain hours of the day considered peak at which much more solar energy is available These are the prime hours for operating the more demanding loads The season has not enabled us to acquire insolation and temperature data for a typical summer San Luis Obispo day so for investigative simulations we use data from a sunny May day in Golden Colorado Figure 5 1 this is the same data used by Den Herder in his simulations 20 Time 0 represents 6 00 AM while time 1439 corresponds to 5 59 AM the following day All daytime plots in this chapter likewise use a minute based time scale starting at 6 00 AM AE y d E A E L 100 200 300 400 500 600 700 800 900 insolation Wm 2 8 time from 6 00 min Figure 5 1 Golden Colorado Insolation and Temperature By way of comparison Figure 5 2 shows insolation and temperature data for a partly cloudy March day in San Luis Obispo typical of the majority of the month Almost 10 66 daylight hours are represented here Peak insolation for these March days appears to be around 420 minutes 13 00 insolation W m42 time from 6 00 min Figure 5 2 San Luis Obispo Insolation and Temperature For the control system it is essential to be able to di
107. which runs on a Linux PC platform The Simulink model is validated by comparison to measured prototype responses Simulations are used to predict SuPER system behavior under various load scenarios in order to maximize battery life The simulation will be a valuable development tool for future SuPER advancements iv Table of Contents O E vii Listiof Table ido ix Chapter 1 Taro UCA AR a ai 1 Ah SUPERS Pro a et lee aso do e O 1 12 Personal OM o dat E a nantes ves 1 1 Sol r Power SS 2 1 4 Thesis Ob VES as 4 1 5 Doc ment AA A toate 5 Chapter 2 Back a o a te as e o O 7 2 L Pha se 0 Prototyper eonen an cade eae ved eR cele ely Ae arent 7 ZA sPOWet and DISTIN asdedaad n E E A aa aa a aan 9 2 1 2 Status and Control Hard War 12 2 2 SUPER Load COTO ad a 15 DDN A A aa a a aa 15 222 A a a a a anal easter ae a a a este 15 J ee AD A EA i E RO 19 22 A LAUSD O 19 2 2 5 DCO MI A G 22 2 3 Battery Management monoica ieni a a a a e E E a week 24 DMA Phaser bap a a Sa tae teas as a ha sues a A SS 26 Chapter 3 Prototype S ida 31 Bed Inter A A e 31 3 2 FUNC HOt A A A 33 A a a a a a e a 36 Chapter 4 MATLAB Simulink Model coi iii 40 BV WOME A tea i a EEE AE Aaa sides ease OS 40 AA A Des iat Approach reroror arian n e evade asaweheas tab a A EE R EAA 42 41 2 Function Blok Siccin nyeni di 46 41 3 DC Motor SUBS SEM aT ER R EA EAR S 55 4 2 Principles of Timing and Samp Wie aiii 58 Chapter 5 Observations and Model AuthenticatiON oooconn
108. y cycle For power efficiency we would clearly like the resolution to be as small as possible but that comes at a cost The duty cycle resolution is the product of the switching frequency fsw and discretized simulation sample time Ts DC step F T Increasing the sample rate comes with the cost of increased simulation time and memory requirements However long sample times can make the duty cycle resolution too coarse to allow a realistic simulation as the converter will be forced to sacrifice large amounts of power Figure 4 18 illustrates the effect that a discretized simulation has on the ability to differentiate between duty cycles The black line is the PWM signal as it would be output from a signal generation block The blue markers are samples spaced at 1 Ts and connecting the markers would represent the PWM signal as it is passed to the MOSFET In this case DCstep the duty cycle resolution has been found to be 5 a duty cycle of 5 in a holds no surprises In b we see that increasing the duty cycle to 9 will in practice be the same as a 5 value We must increase to 10 as c shows in order to the see the change Similarly in the opposite direction d a 1 duty cycle is the same as 5 59 Figure 4 18 PWM Signal Sampling a 5 Duty Cycle b 9 c 10 d 1 In order to prevent the solver from infinitely looping on the m file S function math operations which occurs because of the feedback inherent in

April 2007 (pdf - Cal Poly San Luis Obispo

Contents

Download Pdf Manuals

Related Search

Related Contents