Home

Untitled

1. APPENDIX P SSL ee ee 262120114 ia 3042737183745 2E 04 ET2E12891454 28 D4 BADEIiL43i12EE D2 148435242 E E7038 DE Gr TEE Oh 11327 047288 E34 55476202156 53376 02 ETZJESEDSAIQDSEGE D 1135 52 94497121E D32 ETA4E2171791T15E D4 i01942 9 75 E H XLD3 907135374E DL 4647 DERITDTERi2E Di Sage DEAE GT 12 LED ST 1575747222590 DLE D2 JaAlDINSATSlESE D5 12 2749026759173E D32 SEESTIEJESTOTIE DL PEDE Fae ob 4124171552232 DG E Dri 25415872845 LG ee GED 2S PT O20 IT asye 1944267 18217f 4E D32 ELSE AITE E L 7T20841D idLX4dE Di 123E422T 9TIT25 3631062055 74 19803D84242D475 E D3 X5X1250512534x28 n5 TRG LJ E EITAS Jeana 75 10b EDOL the Fe 1213248352307 1560741256667 7597290593175 32 4h 4 0 X828n85 25XpsE n5 74 20i SE B 275055557755 5 lgEEA4T2TA9El9 201437D742i36i 1005052 40 SoD 3E Oh 9599156872192 05 09 ESETTI2NLT21725E D5 E27 DR Ge Ga Ge 4
2. 48 Mein Graphs load FPE ay VAT Fr AST 0 00 0 10 0 20 0 30 0 40 0 50 0 60 0 70 0 80 Figure 4 11 Sample of Output Voltage for Harmonic Figure 4 11 shows the output current which was measured at the load for 0 8 seconds while Figure 4 12 shows the data generated by PSCAD which is the first column shows the value at particular runtime and the other two columns are the current that flows through the load For this type of disturbance also the length was varied at 50 km 100 km and 200km 0000000000000 10000000000000E 02 20000000000000E 02 30000000000000E 02 40000000000000E 02 bO OO00000000000E 02 60000000000000E 02 0O00000000000E 02 0000000000000 412317054762302E 06 52621950165532E 05 212774304586846E 04 30427371837452 04 37251268 9345432 04 B54063115451253E 02 14854952421947 01 0000000000000 42317054762302 06 52621950165532E 05 212774304586846E 04 3042737183745 27E 04 237251269345 432E 04 54063115481253E 02 14854952421947 01 Figure 4 12 Sample of Data for Harmonic 49 4 2 Software Development The software development can be divided into two parts which are scripts writing and GUI development interfacing The software was developed by using MATLAB environment which are by using wavelet toolbox neural network toolbox
3. Transformer C Transformer D Figure 4 31 Voltage Swell Circuit Figure 4 31 shows the created circuit for voltage swell within PSCAD environment The breaker was applied at the transmission line to produce voltage swell which means increased in voltage for 0 05 seconds Based on data generated by PSCAD it will be used to do wavelet decomposition and then based the Parseval s energy theorem wavelet extraction will be executed Figure 4 32 shows the wavelet decomposition for five levels of Deubechies wavelet for voltage swell at 100 km while Figure 4 33 shows the wavelet extraction of 1t 64 cD1 cD2 cD3 0 5 5 2 0 0 0 0 5 5 5 0 200 400 600 800 0 200 400 600 900 0 200 400 500 900 cD4 cD5 5 5 5 0 0 0 F 5 5 0 200 400 500 900 0 200 400 600 900 0 200 400 600 800 Figure 4 32 Wavelet Decomposition for Voltage Swell at 100 km The level 1 of the detailed coefficients showed that spikes occurred at two different times which also means that swell has occurred for duration of 0 05 seconds The increased voltage during the swell duration causes high amplitude detailed coefficients level 4 cD4 therefore a peak was produced in the extracted feature as the deviation of energy was at its highest by that level The two signals have a high concentration of energy at level
4. 2 P 2 La COREL eh TAE OO OG OH DO 3 24 CHOR DO DER Lr OG OO DO el 3 oe EHEHEHEH E EL e T DIL SL Ee ee ML DIR SS RET ee ed EL HAHH ST EL eee DO DIR Lr
5. i Transmission Line B Transmission Line C ET A B FAULTS ABC 000 100 0 MVA 100 0 MVA er oT T 66 0 1320 132 0 66 0 Transformer Transformer Figure 4 28 Voltage Sag Circuit 61 The circuit in Figure 4 28 was built inside the PSCAD and it 1s able to produce the voltage sag disturbance at the measured voltage The timed fault logic was used to apply fault at the transmission line for duration of 0 05 seconds The data generated by PSCAD will be used to perform the wavelet decomposition and then wavelet extraction Figure 4 29 shows the wavelet extraction for voltage sag at 100 km while Figure 4 30 shows the wavelet extraction from the deviation of calculated energy at each level of coefficients cD1 cD cD3 0 5 5 0 0 0 0 5 5 5 0 200 400 600 900 0 200 400 600 0 200 400 600 900 cD4 cD5 CAS Ca c CT E 5 0 200 400 500 800 0 200 400 600 800 n 200 400 600 800 Figure 4 29 Wavelet Decomposition for Voltage Sag at 100 km Two spikes happened at detailed coefficients level 1 at t 0 2 s and 0 25 s It shows that sag has occurred for durations of 0 05 seconds The other wavelet level has also experienced variations at this same instant At detailed coefficients
6. 86 lload lload2 C 600 gt veo sin th APPENDIX 1 signal for voltage sag 87 APPENDIX J signal for voltage swell 88 APPENDIX K Signal for outage 99 90 APPENDIX L Signal for harmonic i 140300 Dd od nod no ELDEDANELELEDEC ENT MOEA Ore 3501000030000 03 eld DOO DO DOE 02 TODO DO DOCE DG E 02 ed Oe DX D
7. 5 5 5 gi emm mh e 5 mo m0 a e Bl zx am Extracted Feature 5 0 Ca 5 am 10 o 15 0 1 2 3 4 5 Level of Decomposition Voltage Sag at 50 km Percentage of Deviation 50 400 cD4 400 600 900 cD 5 0 5 0 200 400 600 900 cD5 5 0 5 ann Ann Rnn ann Extracted Feature 2 3 Level of Decamposition Voltage Sag at 100 km 96 cD3 5 0 5 0 200 400 600 800 5 0 5 n nn ann f un cD1 0 5 0 0 5 0 200 400 600 900 04 5 0 5 200 400 500 800 Percentage of Deviation 95 cD2 5 0 5 0 200 400 600 900 cD5 5 0 5 0 200 400 600 800 Extracted Feature 2 3 Level of Decomposition Voltage Sag at 200 km cD3 5 0 5 0 200 400 600 900 cAS 5 0 5 0 200 400 500 800 Percentage of Deviation 55 200 200 400 cD4 400 600 600 000 800 cD2 5 0 0 200 400 500 900 05 5 0 5 0 200 400 600 900 Extracted Feature 2 3 Level of Decomposition Voltage Swell at 50 km 98 cD3 5 0 5 0 200 400 600 800 cAS 5 0 5 0 200 400 600 800 swell Percentage of Deviation 95 200 400 600 610 cL Extracted Feature 2 3 Level of Decomposition Voltage Swell at 100 km 5 200 cL am 000 mAs swell 1 00 mU 99 100 coz 0 201
8. 5 525TEGDEILNZ2 Dak oe 13 07295420 251 TOHI 1I T1271012 202 D N 5T3x73 2mL 7K 1 8075k247 2025 06 T YXEWESTETEJ B PEOR aa 12 2K129054X AK 32 2036 638 7367 18 21418w152 E 14 427425 56725 l1g 54471737 23 dick oe Sa 1 ME PHTI B 213212254737 The Ph SDSL TTT Biel EP t 15 Sg TERS 1 amp 5406547642504 2352354024 598 eye eT 23 5 NTEEET LIAS 214 219 020 0620 20 01042422 8000 43 re bo Ppa te GER ATE CaaS 11 714508412427 ere pet 25 27l11832525455 22 17 n5k574 7K 2 5 72724124 376 411 _ AC res Tad 72291715808 75 007 sa 421010234 sll Tee 434 3 42 4 Lees i 41 714142221 X NSZTESQADLISSE 2 6542251654705 T 1zTz sata paw i 7J4 amp 77La7 4477 2007953455141 1 44 284224 EGER 4 5428245 22 4083 amp T7 Adis T LEESmUITEShIA Ji ad Tages tes 5 48125 LELNE 442542443114 1 1712685571491 2025010412405 421417 B GEDHAG DiR ET ETETETT 32 84004 9 TOT eed Tee 7D 2 1178157702515 8 MELITES0 i0 amp ad47R dE 15 2308979594811 34 563700272068 EER ETE TEETE bi i12 71921092 32 amp 5 538722 953034 158027272 Ei LER q 5713222274735 TEM Pere tn end by 15 7107324 2734 lr 4
9. M 0 00 0 10 Figure 4 5 0 20 0 30 Graphs 0 40 0 50 0 60 0 70 0 80 sample of Output Voltage for Voltage Swell As the magnitude voltage increased for duration of 0 05 seconds therefore it shows that voltage swell disturbance occurred and the duration depends on the set time for the breaker to operate 0000000000000 i n00000000000E 02 0000000000000E 02 20000000000000E 02 40000000000000E 02 50000000000000E 02 60000000000000E 02 OOOO0DOOODOOOOE 02 Figure 4 6 0000000000000 12454308203937 35 9085 90369092 60402964182625 660667 8783700 76342279344596 54540980621265 10671414863474 0000000000000 15949874373249 30029000862984 30127264831829 12183925201268 23335700502649 71280606452012 1 2256113776101 0000000000000 34955661693125E 01 bB8795895061085E 01 30275699350796 64422753582432 99677979847245 1 2582158707328 1 3325255262445 sample Data for Voltage Swell Figure 4 6 shows the sample of the output data for the voltage swell The first column shows the time and rest of the columns shows the value of voltage for phase A B and C at that particular time This data file can be easily modified for use in MATLAB The simulation was run at different locations which are at 50 km 100 km and 200 km by varying the transmission line 2 to obtain the voltage swell at different di
10. 1924621123278 Lee Sample data of harmonic A2 17954742502E D4 S type led 228 06 2127749044834 E DI 2042737483745 2E D ETIE12891454 2f D4 BAD 15 326002 136 493524215471 DL ITE E Dl 182708728E EAE 337 02 STZ S000 EE 4 20407424 03 ETAEZLT1731715E D4 521 94 41D342307135374E DL 45 47 BEITDT LR i2E Di BESZITl4TT4 4E DL DTb 57127248 Z E Di EANEZIDETTNIDER DU 14574747222 OE Di 184LDLSSZTSIESEK D5 i1243723026 amp 75173E D SEESXIEJERTTIE DL 0 482417155222 DG ok 2941 39748xX54K E DL eae ee 0 2577D2077744 9R D5 2 E D3 4152882795 427ETE DL T490841DEi2LCadE Di 123E4227872723 bed ed oe GLb 525511502315E53E Dl 1303084321204 7 5 E Dd S8 I25D51253X 2f D5 Ritts reel 02 25T3NLTSLSDSMEE D5 4055026204735 1219298352707 Ged 4G eter TSEZ rI 44 426 02 MIO E S A7 0 ESITE 11221 011 21 262 108E4T2749E174 211427D748i36i 1005052042840 5001588741825 8 05 QESETTINDTS1721E D5 es 27 Ged ET E Dl 1247583703182 eter epee rad 1593073097784 94 95 APPENDIX Q Wavelet Decomposition and Extraction 201 ud B 203 5 5 D jede e au apud BID 0 eu 48 Faji
11. EERE CRE 07 i 07 I DR URS 12 17 _ Rite ed Codd Codd Ce a a SOR Oe D DEI De 02 Bit DO 43 1190 00p0D 000E 01 32000 01 CES DETUR 24 1320000000000 Df 11 ORI De 3 1590000D0DJID OE 71 iB0n00l00 0H EHE 04 ETE mL EDO DRIED D De Ol d d 1520000000 31 a DoD CeO Ce Ce Ce 00 D 0 ZL Oe Ded Ce d a 2 OE 01 04 250000000000 I A 2720 DODGE 01 03 ERO D DRE iL 3D DO Od D CEC DE 11 21 geio JRH 5 HEHEHE C8 RE 01 3 amp 3000J010HO Die 0 97000 CI OG 01 3B DO D D C0 CR C d DIDI OO DE 711 APO EROR i S15 IIe 731 CE d ATOR E 071 JOIDA oe 24 43 1 0X DIDI ORE Am DID d4 3470000 D DID 71 43 rOit E De 43 E 1500030 00 DUE 01 E20 000010 D 0 EHI DG EXE CE I8 11 EEH d d DE 201 Rue DO D EDO i TODIDIDIDOID DE 31 OO DS iz 141 CET CRC EHH 771 APPENDIX O Sample data of outage DIENA Oo E 1972192971011 Ariea T AiD 1 11521072E VL 2 0924625666656 2 37 91442272 08 X 2EIEEI320 D a3 O47 PETIT b TEESE 0427 2093 T gn 72 0 02232 5 5414419117907 T i4dirdidjibdm SNAILS Lk eed
12. Figure3 4 Script Writing 3 2 Wavelet Transform Analysis The Wavelet Toolbox 15 a set of functions built on the MATLAB which provides tools for the analysis and synthesis of signals and images and tools for statistical applications using wavelets and wavelet packets within the environment of MATLAB The Discrete Wavelet Transform was chose for this project using Deubechies as the type of wavelet and wavelet extraction is perform based on Parseval s Theorem approach By using the wavelet decomposition of wave the extracted value and pattern of the disturbances can be obtained In a signal lower frequency is more significant than the higher frequency It 1s because it 1s able to show the signal s identity and in wavelet transform two filters are used which are low pass and high pass filter Therefore the disturbance signal then will be passed through two complimentary filters and emerges as two signals There are five level involved for wavelet decomposition and it is depends on the sampling frequency or the highest cutoff frequency set in the analysis The first level of the wavelet decomposition 1s filtering out the high frequency component and the wavelet then is shifted along the signal in convolution the components in the signal that matched the Wavelet s frequency will be resulting in high amplitude of the coefficient On the second level of the decomposition the frequency of the wavelet is half decreased as the band o
13. f An error dialog This action will close all windows and figures Do you want to quit appeared when the Close option from the File pull down menu of the main windows 15 choose If the users wish to continue with another analysis click and it will return to main windows Otherwise click Yes to exit from the program
14. 90 4 TOL ee ie 41 221108472 LI babe eisai EG cL L iOS LAPSE IE OL 7204017593468 TS LOSS eer Li 214 9975 02452904 24 LEGGEECEME A IE 107 13827213 li B41247220174 y n4E T2NTIN TTD E 5D4n162127 12 1E 18 1225841E28277 24 2z2 9644 B cee 25 253497189 3 20 422417732 27 12 22558Ex 9 9 L TU BOE JL 4 2i147242604162z 17 232 Ap 28 40656 EDO ZE ATIL TOMEJIS Sh 52125 T7 T7D7 TE OL141 3490901 2276923460127 122265 M 21 25220706470 0129042179107 47 RA E 48 11 01 027rT1928 23 fee uid Z2 TIE PE 35 7424E0077327 520220525328 OT ean Tees Tork TI OE eS ee LOPE SS Le ee ET DEL OL Dc SSO eg i A4pELi20i34142 2 2101545 4212142 LA JAHH L E RAISDImIEINEG 40427251643 See 24772512424 28742 P 4T315 47 DE IRE T ANH LIETEN Beet ed ad ed x eS Lao 94212734292 2144 2 4415545414411 17 Se at 10 3662806586200 Ep SLEI rI ANT EA 6287005244 12 D213 1405577 4 245122 4 T2223587522711 3 GEEA 20 257274050215 babe Sete 17 837314751024 2 525 T dd Gm LS S202 567 ED 0169040420 1 24TB342n2 3 ii dLG4r240i06E 2 T742391453 7423 01 LINT A
15. locations which are at 50km 100km and 200km 0000000000000 0000000000000 0000000000000 0000000000000 i 0000000000000E 02 137097740525808 599374725125597E 01 305 3621588580337E 01 20000000000000E 02 29452031860641 40689160287 442 30969362914245 30000000000000E 02 16720925784024 2 2295172590574 53421757230635 40000000000000E 02 323679527864410 907 25431010999 4661754340225 500000000000 00E 02 1 1150107689494 81509378841511 2 0010044635788 60000000000000E 02 2 0434624666616 37 457624048414 2258199660030 O000000000000E 02 2 8804907 668905 40089231266945 9733631894675 Figure 4 9 Sample Data of Outage 47 4 1 4 Harmonic Simulation The harmonic simulation was done by measuring the current across the load as the total harmonic current distortion THDI is more significant if compared by using total current voltage distortion THDV which needs to measure the voltage across the load Thyristor SCR 15 used to create the harmonic disturbance by switching it rapidly and for this circuit 1t use resistor and inductor as the loads SCR Line Switches SCR Load Current Test Point Figure 4 10 Schematic Diagram for Harmonic Disturbance
16. the M File Handles is used to share data between callback and the users can access the data the handles structure in any callback because hObject and handles are input arguments for all the callbacks generated by GUIDE There are two ways to access the template either by entering guide at the MATLAB prompt or 1f GUIDE 15 already open select New from the File menu in the Layout Editor GUIDE provides several templates which are simple examples that users can easily modify it to create their own GUIs The templates are fully functional GUIs and their callbacks are already programmed Users can view the code for these callbacks to see how they work and then modify the callbacks for their own purposes GUIDE provides four templates which are e Blank GUI e GUI with Uicontrols GUI with Axes Menu Modal Question Dialog Besides that command function also was used for this project which 15 inside the M File and the name of the function which defines the first line of M File should be in the same name as the M File It is use for the MATLAB to search when the users try to use the script or function 3l Editor C MATLAB7 work close_all m File Edit Text Cell Tools Debug Desktop Window Help Dae X Ba Boo eb B DB I gt function varargout submenu Callback h eventdata handles varargin function cmd close all handles lt cmd close all
17. OCHO HOO OO OOOO OOOOgoodooogoou ia0000000000000E 02z L245 4506203937 15 9496745735249 34955661695125 01 ZODOODOOOOOOODE 0z 35908590369092 30029000862984 557958595061085E 01 30000000000000E 02 504025964152625 301272645318529 30275699350798 40000000000000E 02 66066787 83700 1218239252012069 6442275355 82432 bODODOOOODOOOOOE Dz 6342279344596 23335700502649 996779795847245 80000000000000E 02 545 40950521265 7128050545 2012 1 2582158707328 ODUOODOOOODODODOE 0 2 10671414565474 i 2256115776101 1 3323255262445 amp ODOODDOOOOODOOE 02 50681033557725 6575642200116 1507555514545 20000000000000E 02 1 2023635198724 1 89231081814751 069073960160268 l DOO0DO0OOOOOOOE U01 T 85418715 77257 1 83927 705 96878 149100350375853E 01 ii000000000000E 01 J2 34512331589211 1 4471615407361 96177505497 i12000000000000E 01 2 4923292238831 72670655 95 70019 1 765 6206251529 13000000000000E 01 2 2724696910075 24835871497 974 2 520808405 9873 L4000000000000E O1 1 6418318633953 1 3469447987394 2 9887766621377 ib gogocooocOoogE 01 647322435236036 2 39835 62451626 3 0456786805230 i6000000000000E 01 5 942769013065 25 3 2183958099371 2 6241267 968718 i7000000000000E 01 1 390945 88207 044 3 6404579596700 7309991353656 i15000000000000E 01 3 0923093494440 3 5461325738950 45382322445094 190000000000 00E 01 3 9363334190942 2 890141047
18. a Voltage Sag Figure 2 1 Sag Voltage sag is a decrease of 10 90 of AC voltage and it usually occurred between the duration of 0 5 cycles to 1 minute s time at a given frequency The common causes for sag are heavy startup loads and remote clearing faults performed by utility equipments b Undervoltage Figure 2 2 Undervoltage Undervoltage 1s also usually known as brownouts and it 1s the result for the longer time problems of voltage sags It is closely related with increase in currents as the motor starting required six times or more than its normal running current It could give effects on incorrect operation of control devices computer system crash speed variation or stopping of motors and so on For example it can create overheating in motors and also could cause to the failure of nonlinear loads such as computer power supplies Voltage Swell Figure2 3 Voltage Swell Voltage swell 15 an increase of AC voltage for 10 90 of rms voltage for duration of 0 5 cycles to 1 minute Usually the sources that could bring to voltage swells are sudden load reductions with a poor damaged regulator high impedance neutral connections and a single phase fault on a three phase system Swells effects 1s able to disturb electric controls and motor drives mainly on adjustable speed drives which can be trip because of their built in protective circuitry d Overvoltage Figure 2 4 Overvoltage Overvoltage is the conseq
19. button You can either select the Run command by Tools gt Run or by pressing Ctrl T LIRE J IM 1 nes EJ sbie ha Rl Vir Laer Teel Hap EB Hee Micros METLOB T XLI LIE OB eme POWER QUALITY DISTURBANCE AND LOCATION DETECTOR USING ANN IMPLEMENTATION OF NEURAL NETWORK IN CLASSIFYING THE TYPE DISTUREBAMCE AND ITS LOCATION Program Feature s and Operating Principle The control panel for loading data plot data wavelet decompose extract and neural identification Data plot graph Load Dosis 8 Wavelet Decomposition Graph of Wavelet Extraction Parohia p Chew a Leni 1522 Ce ie eel Fauna T T t p Sinusoidal zm i a i 2 Lug of Deco cation Voltage Sag and Outage Voltage Swell Harmonic References for wavelet extraction Percentage of Deviation No Disturbance Detected Type of disturbance Detail Function List of the Program a uN 11 11 above window appeared as soon as the user select Start from File drop down menu on the main page From this window users have to select and click on the appropriate button to go through the wavelet analysis progress and neural network identifier List of functions of program main window 1 File drop down menu Contains command such as St
20. four types of disturbance which are voltage sag voltage swell outage and harmonic on three different locations which are at 50 km 100 km and 200 km were shown in appendices ant its clearly showed that the energy deviation produced were different for all three locations but however the form are still the same These three locations can represent short transmission line middle transmission line and long transmission line The data from each wavelet extraction feature then will be implemented for further use in to determine neural network structure 12 In this project probabilistic neural network was chose to classify the types of disturbance and location It is simpler and easier to use compared to backpropagation neural network Figure 4 40 shows the input for the input layer of PNN while refers to the target that will produce by the input matrices Then a network was created and simulated by using the input P to make sure that it does produce the correct classifications Editor C MATLAB work my fyp testpnn m Edit Text Cell Tools Debug Desktop Window Help TELTE d OO BOTs NEM Tg 5 2345 6 769 10 11 12 T indevec Te net newpnniP Tl Figure 4 40 M File for Probabilistic Neural Network Then another M File is created which contains testing command for the designated probabilistic neural network and it 1s used to test either the output produced follows the desired or not In order to determine eithe
21. from the electricity service provider Display the energy graph for each wavelet decomposition on the GUI d Use other types of neural network such as fuzzy neural network backpropogation and etc e Add many types of power quality disturbances such as capacitor switching flicker surge notching and so on 76 REFERENCES Ahmed Mohamed Gaouda 2001 Wavelet Automated Recognition System for Power Quality Monitoring University of Waterloo Ontario Doctor of Philosophy Thesis Ahmed Osman Ahmed 2003 Transmission Lines Protection Techniques Based Wavelet Transform University of Calgary Alberta Doctor if Philosophy Thesis Antony Cozart Parsons 1999 Automated Location of Transient Power Quality Disturbances University of Texas Austin Doctor of Philosophy Thesis Bishop C 1995 Neural Network for Pattern Recognition Oxford University Press Bhaharudeen Ali Ahmed Shahabuddin 2007 Artificial Neural Network for Power Quality Disturbances Recognition Using Wavelet Transform Analysis University Technology of Malaysia Carling A 1992 Introducing Neural Network Wilmslow UK Daniel T Kaplan 2004 Introduction of Scientific Computation and Programming Canada Bill Stenquiest Don Percival 2000 An Introduction to the Wavelet Analysis of Time Series Applied Physics Lab University of Washington Seattle Fausett L 1994 Fundamentals of Neural Networks New York Prentice Hall 10 11
22. level 4 it shows that the highest amplitude occur which means that the signal concentrates most at this level This implies that level 4 carries the frequency that nearly same with the fundamental frequency 50 Hz 62 Level of Decomposition of Energy Deviation 0 003827 Table 4 1 Percentage of Energy Deviation for 5 Levels of Wavelet Decomposition for Voltage Sag at 100 km Extracted Feature Percentage of Deviation 1 0 1 2 3 4 5 Level of Decomposition Figure 4 30 Wavelet Extraction for Voltage Sag at 100 km Based on the extracted feature showed in Figure 4 30 it shows that level 4 has the most percentage of deviation which means that both signals energy have a high concentration of energy at that level However the energy of voltage sag at level 4 1s slightly lower that the energy of reference signal at that level Therefore negative peak deviation was produced at that level The voltage drop during the sag duration causes low amplitude in detailed coefficient level 4 cD4 63 4 3 2 Voltage Swell Transformer A Transmission Line A Transformer B 100 0 k 100 0 MVA T e 100 0 MVA n Tw 180 9 hm el 11 0 132 0 100 0 km 132 0 11 0 1 Transmission Line B Transmission Line C 4 wy 0 001 wy 0 001 wy 0 001 100 0 MVA 1 100 0 MVA 2 660 1320 132 0 660
23. load Figure 4 34 Outage Disturbance Circuit Figure 4 34 shows the outage circuit and the voltage was measured at the load The breaker was put at the load in order to produce momentarily zero voltage The data from PSCAD is used to perform wavelet decomposition The energy at each level will be calculated and the deviation from each level of decomposition of disturbance signal and reference signal then will be used to perform the wavelet extraction Figure 4 35 shows the wavelet coefficients for 5 levels and Figure 4 36 shows the extracted feature of outage at 100 km 67 cD1 m cD2 cD3 0 5 5 5 0 0 0 0 5 5 5 0 200 400 600 900 0 200 400 600 900 0 200 400 500 900 04 E AB 0 200 400 600 O 200 400 600 800 Figure 4 35 Wavelet Decomposition for Outage at 100 km cD5 5 5 5 I 3 5 5 5 800 0 200 400 600 900 Based on Figure 4 35 the detailed coefficients level 1 has two spikes at two different time which equals to 0 05 seconds The energy concentrated at level 4 of detailed coefficients as the amplitude is it highest at that level compared to other level Therefore the deviation of energy for disturbance signal and reference signal produce the highest percentage at that level as shown in Figure 4 36 Level of Decomposition of Energy Deviation 0 040518 0 0103 0 89134 11 9152 _ 0 045697 Table 4 3 Percentage of Energy Deviation for 5 Levels o
24. program is saved into the compact disc CD as a soft copy for future references and development NO Study on Types of Disturbance VVavelet Transform and ANN Design Circuit and Collect Data by PSCad Extract Characteristics by VVavelet Transform Divide Data into Training and Test Set Determine ANN Network Structure Insert Data to Input Layer Training Process Test and Confirmation Achieved Desired Result GUI Development Project Success Figure 3 8 Flow Chart Process 39 40 CHAPTER 4 RESULTS AND ANALYSIS 4 1 Data Simulation The PSCAD EMTDC is used to perform simulation which then will create various types disturbance that 1s going to be studied this project The types of disturbance that 15 constructed within the PSCAD are a b c d Voltage sag Voltage swell Outage Harmonic The data obtained from each simulation 1s used for feature extraction and 1t was saved into a data file dat for easier modification The circuit model for each type of disturbances were created based on online dissertation reference books journals power quality websites and other paper researches 41 4 1 1 Voltage Sag Simulation For the voltage sag the interconnected AC system was chosen and the diagram for the desired voltage sag 1s sho
25. 0 4010 am 000 05 cA 5 0 2 000 Gu Extracted Feature swell 0 Percentage of Deviation 961 N 1 2 3 4 5 Level of Decomposition Voltage Swell at 200 km 101 coz i03 E 3 0 0 0 50 600 0 AD 910 0 zx AUD B 30 Extracted Feature outage Percentage of Deviation 961 0 1 2 3 4 5 Level of Decomposition Outage at 50 km 102 q 40 oD am 40 aw cD5 5 20 40 0 Extracted Feature outage 00 Percentage of Deviation 96 0 1 2 3 4 5 Level of Decomposition Outage at 100 km Percentage of Deviation 96 2 eo 400 a am coe Extracted Feature 2 3 Level of Decomposition Outage at 200 km cra 40 600 putage 00 670 103 Percentage of Deviation 96 nme mes 01 a un ang 90 E 2 5 D Extracted Feature 0 1 2 3 Level of Decomposition Harmonic at 50 km 104 cL 5 D 200 AD BID ED LAS 5 am 800 0 105 cL cDz i03 le 5 E 4 200 610 610 71 200 bp am 0 2007 490 ai 3 i i 4 fi Ani Rn ani n in Enn Enn Extracted Feature harmon IC 00 Percentage of Deviation 961 0 1 2 3 4 5 Level of Decompos
26. 0741 I 0461923720201 20000000000000E 01 4 2691635612924 1 7149161297905 2 5542474315016 21000000000000 01 9543207205032 15162323540831 3 83206974550949 22000000000000E 01 3 0645557775321 1 59350066538549 4 65 80594428970 23000000000000E 01 1 592327596185 5 3 26062405246709 B5 85654205567 243000000000000E 01 dbb24221275286 4 60138267749718 4 3461445622189 25 000000000000E 01 2 2270562 9869354 5 26423549875 932 3 1371786888998 26000000000000E 01 4 0321403567213 5 530175 9511395 1 3580356244176 27000000000000E 01 5 3824793620952 4 65158853847748 7665 9397752544 28000000000000E 01 6 03531265790919 3 0965986700364 2 9417139879556 29000000000000E 01 5 8488449935230 0058682938990 4 8429766996240 30000000000000E 01 4 7816670906035 1 3832183698496 6 1648654004532 31000000000000 E 01 2 9355961689534 3 7335079255336 amp 6691040945 170 32000000000000E 01 533968887 245898 5b 6913604239352 6 2253293111842 33000000000000E 01 2 1016785332231 6 9396250889283 4 8379465557051 34000000000000E 01 4 5945553023414 T 24510 74137177 2 65355 21113763 3b 00000000000E 01 6 566465 6000306 6 b5bi36076350129 52857965017 944E 01 26000000000000E 01 7 6945689811549 4 7828245870282 2 9120442941267 37 D000DO0O00DOODOOE 01 7 645837696640 2 2537277040646 5 51085 6065 5 994 38000000000000E 01 G r 059724947576 4591132888117 7 45 18838236538
27. 1 0 1 0 HOw A Ye B Bay He MEO Ew 110 1320 12 1000 km C 4320 110 ANNECY Transmission Line B Transmission Line C AM oH _ Tim ed Fault gt Logic FAULTS 2 8 8 B8 ZH Qe e og od 3 3 3 5 5 5 gt A 4000IMVA 100 0 Mva Pari Bar 1320 660 A ob oH Transformer D Transformer C oad load load Ea Eb Ec load load 84 APPENDIX F Schematic diagram for voltage swell Transformer A Transmission Line A Transformer B 100 0 k SAA 100 0 M 100 0 1 0 Moon 1 0 HOE E nl Br el B NBEO I SJ amp AN 110 1320 1090 1320 110 16 Transmission Line B Transmission Line C BRK A B s s s T 82848 54 ed C OH Og 3 3 3 3 3 3 aS e deam A 100 0 MVA A A 100 0 MVA A Bay HE Bay C 66 0 132 0 8 C 4320 660 HH Sp H Transformer C Transformer D load load load load load
28. 12 13 14 Wiley 15 16 17 18 19 20 77 Guilermo Miguel Riera Ayala 1996 New Wavelet Indexes for the Severity Assessment of Transient Disturbances on Low Voltage Distribution Systems The George Washington University Doctor of Philosophy Thesis Gurney K 1997 An Introduction to Neural Networks UCL Press Haykin S 1994 Neural Network A Comprehensive Foundation New York Macmillan Publishing Haykin S 1999 Neural Networks a Edition Prentice Hall 1999 J Arrillaga N R Watson S Chen 2000 Power System Quality Assessment Jianguo Liu 2000 Wavelet Modeling of Powr Transients Clarkson University Doctor of Philosophy Thesis John J Benedetton et al 1994 Wavelet Mathematics and Applications Florida CRC Press Inc Corporate Blvd N W Boca Raton John J Grainger William D Stevenson Jr 1994 Power System Analysis United States of America Mc Graw Hill Lee Chung Haw 2006 The Implementation of Wavelet Transform in an Automated Recognition System for Power Quality Disturbances University Technology Malaysia Omer Ozgun 2001 New Models and Simulation Techniques for Power Quality Assessment Texas A amp M University Texas Doctor of Philosophy Thesis Penna C 2000 Detection and Classification of Power Quality Disturbances Using the Wavelet Transform Universidade Federal de Uberlandia Brazil M Sc Dissertation 78 PAR Patterson D 1996 Artificia
29. 21 Power Quality 2 4 1 Types of Disturbance 2 1 2 Power Quality Solution 2 2 Wavelet Transform 22 4 Continuous Wavelet Transform 2 2 2 Discrete Wavelet Transform 2 2 9 The Advantages of Wavelet Transform 2 3 Artificial Neural Network 2 341 Learning Phase 2 3 2 Training Phase 2 3 9 Types of Neural Network 2 3 4 Feedforward Neural Network 2 3 5 Radial Basis Neural Network 2 3 6 Advantages of Neural Network METHODOLOGY 3 1 Software for Project Development 3 1 1 PSCad 3 1 2 MATLAB 3 2 Wavelet Transform Analysis 33 Artificial Neural Network 3 4 Software Implementation and Result Analysis 3 5 Composition of User Manual 10 11 11 13 15 16 19 20 20 21 21 22 23 24 24 25 31 36 38 38 viil RESULTS AND ANALYSIS 4 Data Simulation 4 1 1 Voltage Sag Simulation 4 1 2 Voltage Swell Simulation 4 1 3 Outage Simulation 4 1 4 Harmonic Simulation 4 2 Software Development 4 2 1 Script Writing 42 2 GUI Design 4 2 3 Implementation of Program 4 3 simulation Result Extracted Feature 4 3 Voltage Sag 4 3 2 Voltage Swell 4 3 3 Outage 4 3 4 Harmonic CONCLUSIONS AND RECOMMENDATIONS 5 Conclusions 3 2 Recommendations REFERENCES APPENDICES 40 41 43 45 47 49 49 50 55 60 60 63 66 69 74 75 76 79 1X FIGURE 2 1 2 2 2 3 2 4 253 2 6 23 2 8 2 9 2 10 2 11 2 12 2 13 2 14 2 15 2 16 2 17 3 1 222 3 3 3 4 3 5 LIST OF FIGURES TITLE Sag Undervoltage Voltage S
30. 4 but the energy of voltage swell 15 higher than sinusoidal because of the swell effects Level of Decomposition 9o of Energy Deviation 0 04982 0 6596 20 3558 2 4116 0 4931 Table 4 2 Percentage of Energy Deviation for 5 Levels of Wavelet Decomposition for Voltage Swell at 100 km 65 Extracted Feature swell Q0 Percentage of Deviation 96 0 1 2 3 4 5 Level of Decomposition Figure 4 33 Wavelet Extraction for Voltage Swell at 100 km The pattern for wavelet extraction for voltage swell shows that the percentage of deviation has its highest magnitude at level 3 The wavelet extraction pattern showed that it has met the desired pattern for voltage swell which has a positive peak based on the reference that is used to determine the types of disturbance The data from this feature will be implemented into the neural network to recognize type of disturbance occur and the location 66 4 3 3 Outage Transformer A Transmission Line A Transformer B 100 0 MVA 100 9 km 100 0 MVA 100 0 km T 0 11 0 132 0 100 0 km 1320 11 0 A Ly Transmission Line B Transmission Line C I 0001 wy 0700 0001 000 100 0 MVA 100 0 ayy 2C 66 0 132 0 132 0 66 0 i i Transformer C Transformer D oad
31. 42928395185 1 01668544521951 5000000000000 0E 02 2301501951161 1 5664802926250 2 76356699024911 60000000000000E 02 5777352436350 1096437 2908852 4 687 3789727239 7 0DDD00O00O0OODE 02 7 768381154008 2 5162918222591 6 2931299376599 2 3 4 4 3 Figure 4 3 Sample Data of Voltage Sag Figure 4 3 shows the sample of the output data for the voltage sag The first column shows the time and rest of the columns shows the value of voltage for phase A B and C at that particular time This data file can be easily modified for use in MATLAB The simulation was run at different locations which are at 50 km 100 km and 200 km by varying the length of transmission line 2 to obtain the voltage sag at different distances away from the fault 43 4 1 2 Voltage Swell Simulation For the voltage swell the effect of breaker 1s used to give the swell form in the voltage Source Transformer Transmission Transformer Source Line 1 Transmission Line Transmission Line 2 3 Load Voltage Test Point Fixed Load Fixed Load Figure 4 4 Schematic Diagram for Voltage Swell Disturbance The voltage swell happened as the breaker was set to open at 0 2 second and close at 0 25 second for a duration of 0 05 seconds The fault was applied by opening the 44 breaker which will make the open circuit occurred at this state The load voltages however increased for this period of time This caused the load voltage increased
32. 5 5 0 200 400 500 900 0 200 400 600 800 0 200 400 500 900 04 cD5 cAS 5 5 5 0 0 0 5 5 5 0 200 400 500 900 0 200 400 600 800 0 200 400 600 800 Figure 4 38 Wavelet Decomposition for Harmonic at 100 km Based on Figure 4 38 the energy concentrated at level 4 of detailed coefficients as the amplitude is it highest at that level compared to other level This implies that level 4 carries the frequency that nearly same with the fundamental frequency 50 Hz Level of Decomposition of Energy Deviation 2 1491 13 1125 39 273 58 4899 0 18371 Table 4 3 Percentage of Energy Deviation for 5 Levels of Wavelet Decomposition for Harmonic at 100 km p Extracted Feature harmonic 00 Percentage of Deviation 1 70 1 2 3 4 5 Level of Decomposition Figure 4 39 Wavelet Extraction for Harmonic at 100 km From the extracted feature two peaks occurred which is at level 2 and level 4 of decomposition This implies that percentage of deviation is significant at this level However at level 2 positive peak was produced which means that the disturbance signal holds higher energy than reference signal while at level 4 negative peak occurred which means that the energy of disturbance signal is lower than the reference signal at this level The simulation on PSCAD was done for three distance of transmission line which 1 at 50km 100km and 200km by varying the length of transmission line A The wavelet feature produced on
33. 8 38000000000000E 01 4 6131278447002 3 8006987405829 8 4155265552851 20000000000000E 01 i 7353244266531 6 4651211276520 8 2034555743351 1000000000000E 01 1 5375611452645 6 32847 71922247 6 7 9091560465603 42000000000000E 01 4 75532941635 91 9 07 75920629790 4 3222626446199 43000000000000E 01 7 4418260144251 5 5481220784222 1 1062960639971 44000000000000 E 01 9 1793759248082 7558934883816 2 4234824364266 S5 000 000000000E O1 9 67011345 40291 3 9029103065549 B rG6 720231474742 6000000000000E 01 8 7953767736739 35596430717543 8B 42941246564355 7 O00000000000E 01 6 592167 9502626 3 400171 91026555 9922870531380 45000000000000E 01 3 3517838545369 6 8287012824455 180485166983 490000000000 00E 01 51031891761896 9 4179794684145 B8 9076005507956 b DOODOOOOOOOOOE 01 4 460115 1127072 10 757115715846 6 2970046031334 51000000000000E 01 7 5042 909140516 i0 44108351301535 2 6367 9421614350 52000000000000E 01 10 035727053861 8 6103953741390 1 4253316797222 530000000000 00E 01 10 857578336378 5 5699909025229 5 28758743555549 5b4000000000000E 01 10 154643593417 1 7475394210521 8 4071041723650 55000000000000E 01 B 025G5025437054 2 2200995 767215 10 345749861427 560000000 0000E OL 4 7695085147385 6 0616395331316 i0 831245047870 57000000000000E 01 84381895733490 5 9515936449026 9 79541260022375 DIRIJO LORD CRD ee OD 2 Stee eae
34. ANN such as either ANN will fire or unfire The separating line the will be Equation 2 5 b xw 0 The ANN can be categorized into 3 models which are feedforward feedback and cellular and each model can be either in supervised or unsupervised mode Input layer Hidden layer Output layer Figure 2 15 Feedforward Propagation Input layer Hidden layer Output layer i RD 27 E J Figure 2 16 Feedback Propagation 19 2 31 Learning Phase Learning means using a set of observations that 1 used to solve the task such as to find f eF where F is a class of functions There are three major learning paradigms which are supervised learning unsupervised learning and reinforcement learning The cost function C is used to determine how close the problem from the optimal solution In supervised learning the main objective 1s to gather the mapping that 1s used indirectly by the data The mismatch on cost function contains information about the problem domain and it 1s happened between mapping and the data For unsupervised learning the cost function 15 dependent and be minimized by any function of the data x and the network s output f Meanwhile in reinforcement learning the data 1s regenerated by an agent s interactions with the environment 20 2 3 22 Training Phase In the training phase the model needs to be trained and correct training patterns should be determined so that t
35. FYING THE TYPES OF DISTURBANCE AND ITS LOCATION NOR HASYIMAH BINTI ALI A thesis submitted in fulfillment of the requirements for the award of the degree of Bachelor of Electrical Engineering Electric Faculty of Electrical Engineering Universiti Teknologi Malaysia MAY 2008 I hereby declare that this thesis entitle Implementation of Neural Network in Classifying the Types of Disturbance and its Location is the result of my own research except as cited in the references The thesis has not been accepted for amy degree and is not concurrently submitted in candidature of any other degree ui Signature ib AW Author s Mame NOR HASYIMAH BT ALI Dare It MbY Em z To my beloved parents brothers sisters and all my friend for their support and encouragement ii 1 ACKNOWLEDGEMENT First of all I would like to express my gratitude to Allah the Almighty because with His guidance and blessing I am able to finish this report I also would like to thank you to my supervisor En Alias b Mohd Yusof because without his support and advice I got strength to do as at first I was nearly to give up His help either directly or indirectly had given me a lot of knowledge about this project that I can use it for future use Lastly for my beloved family and friends thank you for encourage m
36. J7T8T8443272E 14 amp SE IAE 45x10324 1714 T GH 51275 6125 p 2222108 5m ii Tb ST ci 17 57012210 34j5 20 4432L 20 SB AELTIE i17 90212i 71 amp 7z 12 71 9 50 27 Caria Dr OE E 2 1459 aga RET eS PA 75115 15 85957 Tam 10 COACH DOO DID 304624284424 Taare 31 GGL GS 4142228 798 2 2258i18 9692020 1 EET ERE CETE d 0702 02046 LS SELLA IB E 12027 200260171 Sn he be ce bd E kT ae 5 551TENIA4EJ71D 3 42444212729c6 10 I2259 705 4E 31 618705108097 m SITs amp Dido d 9255214002148 B TE 234977171 id RTALATiGURLI 15 79244148936 44 808226752042 oe Gate ee 72540 ORE okie PGT m EII2EQETTRTI 13 amp 0085 1350444 15 530 9EIPd12982 18 27917421774LB 15 deid TE 146521 be Pt 1 179L23149 1411 5EQISIDERSGEIM 13 4702203250416 6 19 345242057195 EE Ji ae D 25 112190581217 20 10257278424124 11 387214K31 1 n E Tadd deh TEL ED uP ST eT ee 32 12037 2U l21542510553 24 4 200214234 Im BNrmEELATITE 22 105 20066821 amp 17 257295500718 9 1674249421648 ETE pa EE ER 93 LEED ED DEDE CZ
37. PSE 19 16 Pinar 1 771 UNIVERSITI TEKNOLOGI MALAYSIA DECLARATION OF THESIS UNDERGRADUATE PROJECT PAPER AND COPYRIGHT Author s full name Date al pirh litle Academic Sesion declare thal thi thesis is classlied as i CONFIDENTIAL Contains confidential information under the Official Secret Act 1972 Eu RESTRICTED tesincied intfesmmalion ai specified by the organisation where research was OPEN ACCESS i agree that my thesis fo be published s online open access Fill acknowledged that Unlversili Teknologi Malaysia reserves the right as follows V The thesis is the property of Universiti Teknologi Malaysia 2 The Library of Universiti Teknologi Malaysia has the righ to make copies for the purpose of research only 3 The Library has the right 19 make copies of the thesis for academic exchange T a Certified Dy SIGNATURE OF SUPERVISOR SIGNATURE 13035134 NEW IC NO PASSPORT NO NAME OF SUPERVISOR NOTES tthe thesis is CONFIDENTIAL RESTRICTED please attach with the letter trom Ihe organioton with period and rectors for ar restriction hereby declare that 1 have read this report and in my opinion this report has fulfils the scope and quality for the award of the degree of Bachelor of Electrical Engineering Electric Name of 5 BIN MOHD YUSOF Date 7 s IMPLEMENTATION OF NEURAL NETWORK IN CLASSI
38. R 00 WIDGET DI HIE Da DO DIO OD LEOTE Oe i200 D Ce e Ox Dri 135030 DO DO DO OO DID CD CRORE Dd i 15002 DG DO OC E DOO DO CHORO E Dd LTEMA PIEI Cee Cd 1300 DID DODGE OL TOU OD DEL ER it DI 220000 ODDO DOR 2300 DODO CE DGE DT 2 0 DD Ce EX Oecd Dr TE DGIO EI DO DIE aEDODODOIDO ODE DE 2r0IDGOQOOOIDIE 01 D Ged EROR Crd 29080 0o DO Dd OO 0 E D i 33 DOO DO CERRO E Di wae CeCe Od Ce D SEEDS DOG DIL 0 NT CECE RD ET DE SR Ce Ce DOO DOR Dri SED CED CORI DGE DE DE RD Dr i SL OGRE NIE DE AZ CEU ODIO ODE DE 3 OCR DOO DE Od Ee OR Dr DED RRR E AE DID OU CED CREE Dd i CLIE Dri DRE ELI OR DR Di TEIENEI L DR ER it 5nd D D DEI I DE ADI DOO DOE Dri SS DT AG DID Oe Ce ee Dr T THE OIE 01 DE 532 0910300000 01 APPENDIX M sample data of voltage sag 1523172227317 53 6497430 Pe re ee 1 4 FTL OES LLL ETET 2 7 RESELL LS Lo dpa Bel 13 0 90743
39. TLETA 15 12 amp 6174 1244 i7 B amp z4444iiin 3 7 3453231427722 li amp Rlibdida 28 2 45547278 14 47 amp u 57 E242 2 203704557 25 In 9552n770211 E CHL 2 28 134717122207 ii 2204822412581 41 zp Ek4sEnds17 42 37 542412535415 24 171855479611154 1 2 273007087 17 75 14757002 11 AE STJ 1974813 TELIT 10 ITRE lt i Sdn d ice dik E 7ESERENULLSII 4 ee 2 nini o rea 3 i pra mm pr rrn 2 487 GD od Ped zd Li wit heed tT THI BAAS be i pi a La ST 13 225904140861 TOT Looe i 5 m181264772L 16 3462 iiaa 21 280204274721 24 064011 05381 03 SENS IL Sees 1 23748 S TEL eLOET LET 3D 17022320 7224 TE XT13ETISDAES dz BE 4L 0326419 28 240744 PED D I5 Drzli ngEdg52 ett beet fy T 4E2 2752 7288 9 Dp427TEg2 389734 37 594 9422002287 53 5897 21 032215 1470774 58 12125092704 1 203983225733 AB Ja 24 OE 2 24313 2E22 373 26014 Gs GATS ET SULET S22 DIRE nd bee le wal GL BSL To 9 APPENDIX N sample data of voltage swell OOOOO0o00008000 OCHO CHO OOH
40. al Page of the Program The Contents of the Program Loading the Data into the Listbox Signal Plotting for the Input Data Wavelet Decomposition and Coefficients Plotting Pop up Figure for the Extracted Feature 34 37 39 41 42 42 43 44 44 45 46 46 47 48 48 5 5 52 53 54 54 55 56 56 57 57 58 58 59 4 27 4 28 4 29 4 30 4 31 4 32 4 33 4 34 4 35 4 36 4 37 4 38 4 39 4 40 4 41 Pop up Figure for Neural Classifier Voltage Sag Circuit Wavelet Decomposition for Voltage Sag at 100 km Wavelet Extraction for Voltage Sag at 100 km Voltage Swell Circuit Wavelet Decomposition for Voltage Swell at 100 km Wavelet Extraction for Voltage Swell at 100 km Outage Circuit Wavelet Decomposition for Outage at 100 km Wavelet Extraction for Outage at 100 km Harmonic Circuit Wavelet Decomposition for Harmonic at 100 km Wavelet Extraction for Harmonic at 100 km M File for Probabilistic Neural Network The Probabilistic Neural Network Architecture 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 Xll APPENDIX UJ WO moo cem qoe geo See dup depo G LIST OF APPENDICES TITLE Characteristics of sag swell interruptions Characteristics of DC offset harmonic Characteristics of voltage imbalance notching noise voltage fluctuation power frequency variations Library of PSCAD Schematic diagram for voltage sag Schematic diagram for voltage swell Schematic diagram for outage Schemat
41. al and reference signal was calculated to perform the wavelet extraction and the graph for the chosen type of disturbance can be obtained by clicking on the Wavelet Extraction button and a new pop up figure will appear 59 rage ol Deri min 96 art prrrpearertamm HIN MA Sinusoidal Voltage Sag Yoltage Swell Harmon and Figure 4 26 Pop up Figure for Extracted Feature Finally neural network will perform classification on types of disturbance and the location of it when users select the Neural Classifier button and again a new pop up figure appeared and display the result m ident firatin Fia Lispi od Doramnpaainan ol Duviutibn Da EHH 2 131125 2 38 272 Daad 534253 1E371 ji Harmonic 100km Figure 4 27 Pop up Figure for Neural Classifier Both of Extracted Feature and Neural Identification windows provides option for users to print out the window s display 60 4 3 Simulation Result Extracted Feature In this section the extracted feature for four types of disturbance that already been studied will be discussed by performing the feature extraction analysis 4 31 Voltage Sag Transformer A Transmission Line A Transformer B 100 0 km 100 0 MVA 100 0 MVA 100 0 km T 7 el 110 132 0 mE 1320 11 0
42. and also GUI builder 4 2 1 Script Writing M file is used for the script writing which contains commands for function that we would like to execute In order to create a program M file 1s very important because the name that contains different command 1s used as the callback The first script that is used in this Power Quality Disturbance and Location Detector using ANN 15 for function of loading and manipulating data which was previously generated by PSCAD and saved as dat Command functions Load and Plot were used in performing those desired functions Then the effort is aimed in Wavelet Decomposition analysis for the selected input signal As the command line was chose to execute the Discrete Wavelet Transform each of the command of the Wavelet toolbox performs various tasks required for Wavelet Analysis in signal decomposition and Wavelet coefficients plotting Function Wavedec has been used to perform the decomposition towards the input signal Detcoef and Appcoef were used to determine the coefficients of each decomposition level 50 Finally the Neural Network command lines are created M File As the Probabilistic Neural Network was chosen for the classification of types of disturbance and location therefore command newpnn ind2vec and vec2ind were used to design the neural network The function newpnn creates a two layer network with the first layer contains radbas neurons Command ind2vec converts indice
43. aneous momentary temporary and sustained which each of them can be classified according to the time taken for the situation to happen and heal 10 2 1 3 Power Quality Solutions a Uninterruptable Power Supply UPS UPS 1s a device that is put between power supply and a device to protect devices from disturbance that could directly affect the performance of the device There are three types of UPS that commonly used which are Standby Power Supply SPS hybrid or ferroresonent UPS system and online UPS The UPS has internal battery which also acts as an emergency power supply that could be as a backup power supply when power failure happened This feature 1s very important especially to the computer as the users can prepare to shutdown the computer when there 1s no power supply so that the data 1s not lost due to the power loss b Power Conditioner Power conditioner 1s an electrical device that improves the power quality and it 1s used to provide AC power supply that is free from disturbance to sensitive electrical equipments It 1 also used between the device and power supply such as wall outlet and it is able to protect the equipments from surge brownout noise and other power quality problems Transient Voltage Surge Suppression TVSS TVSS is used to protect sensitive electrical equipments from harmful surge energy and it is a voltage sensitive switch which monitors the AC voltage input and output waveforms It 1
44. art Close Reset Help drop down menu Contains command such as tutorial using program and info about the creator of the program 10 11 Load Disturbance Data Allow the program to load the list of disturbance data file File list The list of data that appear when Load Disturbance Data was clicked Plot Signal Program will plot the signal and pure sinusoidal Wavelet Decompose Program will tart the wavelet decomposition Wavelet Extraction It will allow user to view the extracted feature of the wavelet decomposition Neural Classifier It will perform neural network to identify types of disturbance and location Disturbance Signal The waveform of disturbance signal Normal Signal The waveform of pure sinusoidal Wavelet Decomposition Graph The graph of wavelet decomposition 10 Note for Advanced User The Power Quality Disturbance and Location Detector Using ANN permit users to load their own data signals and perform the simulation and analysis Users are encouraged to compile and store the data a specific folder Also user needs to save their data with extension dat file Before running the program they need to set the data folder as MATLAB current directory or place it into the MATLAB current directory While running the program once the button Load Disturbance Data inside the program is clicked the user data will automatically be loaded into the list box Then user can perform the si
45. bance classify the types and location for every each and lastly developing the graphical user friendly interface for the user easier to use the program The software used for this project is l PSCAD 2 MATLAB PSCAD 15 used to design circuit that 1s able create disturbances characteristics such as voltage sag voltage swell outage harmonics and capacitor switching The data that was generated from each signal that was produced then will be analyzed for further use in MATLAB By using MATLAB the signal created before then will be characterized using wavelet transform analysis and a few data then will be obtained which is very useful to determine the input layer for neural network to be able to classify the location and types of disturbance that occurred Then a user friendly approach will be created to interface the user with the program Software for Project Development PSCAD E m m E i E TIT reri Figure 3 1 PSCAD Software PSCAD Power System CAD provides features for the user to schematically construct a circuit run a simulation analyze the results and manage the data in a completely integrated graphical environment It is a powerful and flexible graphical user interface to the EMTDC solution engine EMTDC is most suitable software that can be used for simulating the time domain instantaneous responses which is also known as electromagnetic transients of electrical s
46. blems happen such as blackout some company might ask for compensation as this situation will cause their machine to lose synchronization and this will bring them to lots of losses Residential consumers also will have less faith in them if this problem of poor services occurred frequently and if there are many choices of power qualities they will prefer to change to the best supplier Loads are interconnected in power grid Thus any disturbance caused by load at other location will give impact on the stability of the power system This requires the power supply utilities to be more concern on the location of where disturbance originated as less times taken to detect less cost will be needed In order to have high accuracy of detection therefore neural network 15 introduce in giving more faster and simpler method in this new advanced technology 1 3 Project Objectives The main objective of this project is to create a program that is able to classify the types of disturbance and its location using Artificial Neural Network and Wavelet Transform Analysis while the general objectives of this project are 1 To utilize ANN technique in order to detect type of disturbance that occurred and its location 2 To develop a user friendly guide as an interface for user to the created software by using MATLAB 3 To create characteristics for each type of disturbance by using Wavelet Transform 1 4 Project Scope Based on the classification
47. e end of this project It contains the wavelet transform and neural classifier which was created before in MATLAB toolbox by using command line The GUI builder was used to compiled all the scripts written before and also could help users to easily used this program Beside that the manual for the created program was also composed to simply help the users so that they can refer it in order to use the 75 program correctly However it still has its weakness as the scope for this project actually focused on five types of disturbances which are voltage sag voltage swell outage harmonic and capacitor switching Nevertheless due to problems faced during the circuit design within PSCad environment the output did not match as desired This project also involves the usage of PSCAD to create model circuits which are able to produce disturbances at different location All the simulations imply the same result in the wavelet process showing that the program can actually be adapted to the real situation case 5 2 Recommendations The project still has its weakness and therefore improvement is very important so that it can meet its best performance For future works some recommendations have been listed based on the problems in order to improve it a Develop an algorithm that can show difference on extracted feature for different location at the same time b Build up a neural network classification based on the original disturbance data
48. e want to train the data and the more we train it the result will be more nearer to our desired target Below is an example of backpropagation for training phase net trainParam show 50 net trainParam lr 0 05 net trainParam epochs 300 net trainParam goal 1e 5 Learning rate should be chose wisely as if the value is too big or too small will affect the desired output Epochs refer to how many times we train the data to reach the target stated while goal 1s used to minimize the error from the output and our target 37 Input Radial Basis Layer Competitive Layer Where number of elements input vector al radbas IWi1 p 217 a2 compet LW21 a1 a1 is element of where Wiis a vector made of the th row of Win Figure 3 7 Probabilistic Neural Network Q number of input pairs number of neurons in layer 1 number of classes of input data number of neurons in layer 2 Figure 3 6 shows the network architecture for probabilistic neural network which consists of input hidden layer and output Each input vector 1s associated with one of target classes From the disturbance data from PSCad the neuron in hidden layer will be trained Then Discrete Wavelet Transform analysis was applied and the energy deviation between disturbance signal and reference signal which is pure sinusoidal will be calculated The data extracted from the wavelet transform will be compared with t
49. e while doing this report because with you guys I feel like I have many supporters that are still with me to continue this project ABSTRACT Power quality 1s become a popular issue in the power industry since the late 1980 s It can cause the electrical equipments to malfunctions and it also could effect on total power lost for days Therefore this project 1s going to introduce a new way which could identify the location and type of disturbance in a faster and simpler way by using automated recognition system using software The approach used for development of this project is Wavelet Transform Analysis and Artificial Neural Network From PSCAD the circuits for disturbances were created according to their types which are voltage sag voltage swell capacitor switching harmonic and outage The simulation results will show that each of disturbances will produce different waveforms from the pure sinusoidal and therefore Wavelet Transform will be used to produce the characteristics of the disturbance Then ANN will be applied to detect the type of disturbance and also its location in the transmission lines Finally a program named Power Quality Disturbance and Location Detector Using ANN was introduced which include involvement of Wavelet Toolbox Neural Network Toolbox and GUI builder from MATLAB vi ABSTRAK Kualiti kuasa telah menjadi isu popular di dalam industri kuasa semenjak tahun 1980 Ia boleh menyebabkan peralatan elektrik me
50. ed time index ot sampling interval o angular frequency 13 2 2 2 Discrete Wavelet Transform CWT is a costly process and most of the signal data were stored in discrete in computer to process The low pass digital filter 1s the scaling function of the Wavelet g n while high pass digital filter is wavelet function of the Wavelet h n The low frequency content is very significant as it can give the characteristics of the signal for many signals cD High Frequency PRED re 5 Low Frequency EHOD AAA Figure 2 7 Sub band Filtering of DWT The signal will pass through two filters which are low pass filter and high pass filter and it will produce two other signals which are details and approximation Details are the low scale high frequency components while approximations are high scale low frequency component The signals also will be down sampling by eliminating half of the signal when it passes through the filters in order to maintain the number of data we had in our signals This is indicated by Figure 2 7 14 Figure 2 8 Wavelet Decomposition Process For multiresolution analysis using DWT process a signal 1s decaying into a discrete number of arithmetic frequency bands In the first level of decomposition the entire high frequency is filtered out The Wavelet is shifted along the signal convolution and all the component in the signal that matched the Wavelet s freque
51. f Wavelet Decomposition for Voltage Outage at 100 km 68 Extracted Feature outage 1 00 Percentage of Deviation 96 0 1 2 3 4 5 Level af Decomposition Figure 4 36 Wavelet Extraction for Outage at 100 km Negative peak was occurred then at level 4 as the energy of outage is slightly lower than reference signal From the extracted feature it also shows that level 4 has the most percentage of deviation which means that both signals energy have a high concentration of energy at that level The decomposition of outage appears almost as the same of voltage sag Therefore neural network will be implemented to classify wisely types of disturbance that occurred by choosing the data correctly 69 4 3 4 Harmonic lt ia ie s9zo dwyo 0701 s9zo dwyo 0701 5 5 lload2 pos th CO 600 5 vwo S sin th Figure 4 37 Harmonic Disturbance Circuit The switching on and off of SCR produced the harmonic and the current across the load was measured as THDI is more significant in showing the effect of harmonic compared to THDV Figure 4 38 shows the wavelet decomposition for five levels of Deubechies wavelet for voltage swell at 100 km while Figure 4 39 shows the wavelet extraction of it cD1 cD cD3 5 5 0 0 0
52. f the signal now 1s half of the original signal 22 The amplitude of the coefficients increased with the increasing of the level of Wavelet decomposition until it reaches certain level when it started to decrease The drop of coefficient s amplitude implies that the signal is completely decomposed Figure3 5 Wavelet Decomposition Tree The approach developed for the wavelet extraction 15 based on the study made by Penna C Detection and Classification of Power Quality Disturbances using the Wavelet Transform in year 2000 By using DWT and certain features of the decomposition levels of a signal were observed some important conclusions could be made This information is able to detect locate and classify the types of disturbance A digital program was created and implemented in the Wavelet toolbox of the MATLAB platform through five steps as follows Step 1 Evaluation of the wavelet coefficients of the signal in study The disturbance signal is decomposed by Wavelet transform in MATLAB into several levels and the coefficient of each level will be concerned 33 Step 2 Evaluation of the square of the wavelet coefficients found at step 1 Step 3 Calculation of the distorted signal energy in each wavelet coefficient level The energy mentioned above is based on the Parseval s theorem the energy that a time domain function contains is equal to the sum of all energy concentrated in the different resolution levels of the corres
53. gure 2 10 Deubechies Wavelet Basis Functions 16 2 3 Artificial Neural Network ANN Artificial neural network is a model that was created based on biological brain and it consists of neurons that were interconnected with each other These processing signals produced output from input signals it received to solve certain problems Therefore the input signals will be processed and then from the tasks that have been trained output will be produced Figure 2 11 Figure 2 12 An Artificial Neuron 17 There are a few terms that have been used to explain how to implement this ANN Figure 2 13 above showed an artificial neuron that is used in neural network Each neuron are activated by activation functions and for each interconnection there is an associated weight which is Wji The signals sent then will be processed by multiplying the activation signals with its weight Figure 2 13 Artificial Neural Network some ANN used bias signals b and it acts same with weights The only difference 15 that bias will always have activation equals to 1 This connection will be treated equally with weights and then will be adapted according to the learning rule of ANN It can increase the signal levels which will also improve the union Neural network input S b Figure 2 14 Threshold 18 The threshold is quite similar with the bias but it is not adapted and its value will be used to make decisions in
54. he implemented data used inside the probabilistic neural network structure and then the output will be recognized 38 3 4 Software Implementation and Result Analysis When the Power Quality Disturbance and Location Detector Using ANN was successfully developed and functions as desired it is used then by based on the data from circuit simulation from the PSCad The simulation done comes from different types of disturbance such as voltage sag voltage swell harmonic outage and capacitor switching and besides that it was done on different locations which are at 50 km 100 km and 200 km The results then will be implemented towards the extracted feature which will be analyzed and evaluated 3 5 Composition of User Manual A user manual is created 1n order for the user to use the program easily and it includes the installation and setup requirements and procedures explanation of the basic function the features of Power Quality Disturbance and Location Detector Using ANN and examples of using the whole program Besides that a table of troubleshooting concerning on the simulation program is discovered as well The manual for the program are compiled using Microsoft Word 2003 and all the files then are saved in PDF files Users can also access the tutorial that already provided inside the program itself which can be assessed when running the program All the tutorial files including the programming and source code of the
55. hey can perform the correct task Here an incorrect selection which is wrong training patterns will produce wrong results There are four initializations that need to be considered in this training phase which are a Choose correct number of iterations stopping criteria b Learning parameter momentum parameter c Randomize the weight correctly d Minimum error stopping criteria 2 3 3 Types of Neural Network There are many types of neural network which are a Feedforward neural network b Radial Basis Function RBF network c Kohonen self organizing network d Recurrent neural network 1 simple recurrent network Hopfield network 111 Echo state network IV Long short term memory network e Stochastic neural network 1 Boltzmann machine f Modular neural network 1 Committee of machine Associative neural network ASNN 21 2 3 4 Feedforward Neural Network The feedforward neural network is the first and simplest method for artificial neural network In this network the information moves to the forward where is from the input nodes to the hidden layer and lastly to the output The information that was sent did not creating loops in the network 2 3 5 Radial Basis Neural Network i 1 Outputs l Output Layer Hidden Layer Input Layer Figure 2 17 Radial Basis Function RBF Network The network consists three layers which are input layer hidden layer and output layer This netw
56. ic diagram for harmonic Signal for voltage sag Signal for voltage swell Signal for outage Signal for harmonic Sample data for voltage sag Sample data for voltage swell Sample data for outage Sample data for harmonic Wavelet Decomposition and Extraction User Manual PAGE 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 108 xiil CHAPTER 1 INTRODUCTION 1 1 Project Background The usage of electronic equipments 15 important in our lives nowadays and the development of this technology is growing rapidly days by days This newer generation load equipment was created to be more sensitive than the electronic devices before and this will increase the power quality problems As things now are interconnected in a power grid therefore disturbance from load will give huge impact on the power system that will brings side effects to the other consumers too Power quality can be defined as any power problem manifested in voltage current or frequency deviation that can results in failure or misoperation of customer equipment So it also means that power quality problems can cause the load to malfunction as the load was connected to the power system supplied by power utilities Although several methods had been introduced to detect this disturbance but it 1s expensive and took a long time to solve it Time is very important because the longer time taken the higher the affects on power system Suitable proble
57. ile initializes the GUI and contains a set of outline for all the GUI callbacks the commands that are executed when a user clicks a GUI component Using the M file editor users can add code to the callbacks to perform the functions they want them to All the programs created before was compiled using the GUI for users to access simulate and can simply modify their data whenever they want So anyone without the knowledge of the wavelet transform and neural network can handle the program easily However before designing the GUI the programmer should know the detail features of the GUIDE For example each button has its own callback which 1s a function definition line and by using this function we can relate the button with the command that we created before in M file So the button will function according to our desire by adding the code that we already specified as the name of our M file Below is the example of function callbacks that is used for this project Executes on button press in pushbutton3 function pushbutton3 Callback hObject eventdata handles decompose handles Executes on button press in pushbutton4 function pushbutton4 Callback hObject eventdata handles extraction handles 30 As the users press the button with tag named pushbutton 3 then the GUIDE automatically will be linked with the M Files named decompose Therefore the program created will functions as the desired scripts writing
58. ing fundamental sinusoidal wave signal at each Wavelet transform level edis i Energy distribution concentrated in each Wavelet transform level of the disturbance signal in study eref i Energy distribution concentrated in each Wavelet transform level of the correspondent fundamental component of the signal in study eref 4 Energy concentrated at the level 4 which concentrates the highest energy of the corresponding fundamental of the signal in study The energy deviations ed i curve against level of decomposition is then plotted The deviation curve of every particular power quality disturbance has unique features that can be used to identify the problem in the voltage waveform 36 3 3 Artificial Neural Network There are many types of neural network and for this project probabilistic neural network was chosen which is fall under type of radial basis neural network Basically the network can either be in single layer or in multiple layers Multiple layers has its own advantage which it 1s able to give more accurate result as different function can be used for the hidden layer and output layer Probabilistic neural network was chose because it is easier than backpropagation neural network and besides that the result 1s more accurate In probabilistic neural network we need to classify the group for each group of data that we defined Compared to backpropagation we need to determine the learning rate and how many times w
59. irst time they are encouraged to read the manual provided first First of all they need to copy the whole program into the MATLAB folder and placed it inside work file The location of MATLAB folder depends on where they locate it during installations Then they can start the MATLAB window and choose to new GUI from the file menu After that they need to click on open existing GUI and browse the file named interface 1 A GUI window appeared and to activate it they can choose run from the layout editor toolbar Figure 4 20 will come out as soon as the run button was click 56 POWER QUALITY DISTURBANCE AND LOCATION DETECTOR USING ANN IMPLEMENTATION OF NEURAL NETWORK IN CLASSIFYING THE TYPE OF DISTURBANCE AND ITS LOCATION Vrai De corpo eon bari ras 15 de Figure 4 21 The Tutorial Page of the Program 57 ALITY OS TEATRO ETIAM USING ANH mE EE EN EN Figure 4 22 The Contents of the Program Haural Ch esi Figure 4 23 Loading Data into the Listbox By pressing on Plot Signal button the data selected then was plotted on the Disturbance Signal and Reference Signal shows the signal without disturbance 58 Figure 4 24 Signal Plotting for the Input Data Figure 4 25 Wavelet Decomposition and Coefficients Plotting Then the deviation energy from the disturbance sign
60. it can also affect on other end load users So it 1s important to everyone to give concern about this problem From a survey carried out by the Electric Power Research Institute 80 of all disturbances come from inside the homes and business Therefore most of the disturbance comes from household equipments and appliances such as refrigerator air conditioner vacuum and so on Weather conditions like lightning storms and high winds are also another factors for disturbance to occur This power quality problem could give a big impact on economic value either to power supply utilities residential consumers or industry For power supply utilities it will effect on gaining the level of confidence from their consumers and this also will reflect on their number of customers to competing power supplier Residential consumers now for example usually depend on internet for making their business easier either for paying bills online shopping or to communicate with their family and friends which lives far away from them So without computer it will make their life quite difficult as they need to go to the office to settle down their needs 2 1 1 Types of Disturbance There are different types of disturbance that usually occurred in power system such as overvoltage undervoltage harmonic voltage sag voltage swell and so on that can be categorized based on their length and magnitude Each type can be differentiating by its waveform
61. ition Harmonic at 100 km 200 Percentage of Deviation 261 200 cD1 400 cD4 400 600 ann 800 cD 0 200 400 600 900 cD5 5 0 5 ni ann Ann hh nnn Extracted Feature 2 3 Level of Decomposition Harmonic at 200 km 106 cD3 5 0 5 0 200 400 600 800 5 0 5 n nn Ann inn nnn harmonic 00 201 am Percentage of Deviation 551 600 Extracted Feature 2 3 Level of Decomposition Sinusoidal 107 cr E 0 20 6 amp 0 mAb 5 108 APPENDIX R USER MANUAL POWER QUALITY DISTURBANCE AND LOCATION DETECTOR USING ANN USER MANUAL FOR POWER QUALITY DISTURBANCE AND LOCATION 99 DETECTOR USING ANN CONTENTS SECTION D ra Introduction Installation and Requirements Installation of Program Program Feature and Operating Scheme Note for Advanced User Troubleshooting PAGE 11 12 109 INTRODUCTION Welcome to Power Quality Disturbance and Location Detector using ANN the user friendly Neural Network based on Wavelet Analysis program for classification of types of disturbance and location Using MATLAB operation system and the script files along with the GUI builder Quality Disturbance and Location Detector using ANN allows wide analysis of the voltage disturbance data This program requires correct installation of MATLAB version and all the needed script files for program
62. l Neural Network Singapore Prentice Hall 22 Rafael A Flores 2003 State of the Art in the Classification of Power Quality Events An Overview Chalmers University of Technology Gothenburg Sweden 25 Reinhold Rudenburg 1950 Transient Performance of Electric Power System McGraw Hill Book Company Inc 23 Resende J W Chaves M L R Penna C 2001 Identification of Power Quality Disturbances using the MATLAB Wavelet Transform Toolbox Univesidade Federal De Uberlandia MG Brazil 24 Robi Polikar 2005 The Wavelet Tutorial Online Tutorial Rowan University College of Engineering Web Servers http users rowan edu polikar W AVELETS WTTutorial html 23 Roger Dugan Mark McGranaghan Surya Santoso Wayne Beaty 2002 Electrical Power System Quality United States of America McGraw Hill 26 Scott Grimsley 2001 Detection and Measurement of Power Quality Bachelor of Engineering James Cook University 21 Soo Fui Lan ANN to Identify Location and Type of Fault Based on Voltage Sags Fakulti Kejuruteraan Elektrik Universiti Teknologi Malaysia April 2006 29 Shyh Jier Huang et al 1998 Application of Wavelets to Classify Power System Disturbances Department of Electrical Engineering National Cheng Kung University Tainan 70101 Taiwan ROC 30 Tarek Abdelgralil 2003 Automated Recognition System for Power Qualities Disturbances University of Waterloo Ontario Doctor of Philosophy 31 Rip
63. ley B D 1996 Pattern Recognition and Neural Networks Cambridge University Press APPENDIX A Characteristic of Voltage Sag Voltage Swell and Interruption Shon Duration ii Typical voltage Magnitude voltage variation in Electric Power c 220 m Remate System Faults large loads and non linear loads for short duration Ferroresonance Transformers Energy storage technologies UPS 79 80 APPENDIX B Characteristic of DC Offzet Harmonie and Inter Harmonics Waveform pix Typical voltage Magnitude Distortion Non Linear Loads Nonlimearloads System Resonance Resonance C eat Active and Passive Filters Transformers 81 APPENDIX C Characteristics of Voltage Imbalance Notching Noise Voltage Fluctuations and Power Frequency Variation Typical voltage Magnitude Content ae a _ EE A NE EN HE NN 25 Hz 82 APPENDIX D Library of PSCAD Sources Transtomers Breakers Faults Cables Machines P IO Dewces Sequencer Other Single Line APPENDIX E Schematic diagram for voltage sag Transformer A Transmission Line A Transformer B 83 100 0 k SAW AY 100 0 t 90 m 100 0 Mva 4 ANN E
64. load 7 7 Timed 7 2 BEk Breaker Logic o le Eb de Ec APPENDIX G Schematic diagram for outage Transformer A Transmission Line A Transformer B 100 0 k 9 100 0 MVA M 100 0 Mva MAACO 1 0 1 0 100 0 k II CS BAA nl fw 00 0 km ANB 100 0 S 11 0 132 0 0 1320 11 0 AAA Transmission Line B Transmission Line C S 8 od old od e o 5 2 3 3 3 4 On x x 3 3 3 100 0 MVA a 100 0 MVA Br B Br B 660 1320 1320 660 2 lt m i i i i Transformer C Transformer D lt Sh Est Sh load load load load load Tim ed v LAC BRK Breaker Logic 85 load APPENDIX H Schematic diagram for harmonic e 4 m 2 JC A E 8 PE ee L2N a N 3 VCO o 973 H T gt gue VCO L o H m at gt L vco id Alpha VCO D Alpha2 F D e o F o
65. m solving should be taken to restore the system as soon as possible to avoid it from become worse which could affect on total loss of power for days The usage of artificial neural network is popular these days and it 15 widely used in different field as its ability to provide a simpler and faster way than what we already used now It is computer software that 1s under artificial intelligence which also includes fuzzy logic This neural network is easy to use and it also has lots of advantages 1f compared to other intelligent methods 1 2 Problem Statement The intelligent technology that 1s created lately demands power that 1s free from interruption or disturbance However a recent study carried out in USA shows that business sectors losing a lot sum of money every year due to this power quality problems For an industrial sector that used 24 hours machine to operate this phenomenon will give them big impacts especially for the machines that required a long start up time The common causes of this problem can be divided into two groups which are internal and external causes which include weather conditions such as lightning ice storm and wind heavy startup loads faulty distribution components and major switching operations This disturbance can cause on economic value which for power supply utilities they will have to bear the loss that comes from consumers and also from their own power system Besides that when power quality pro
66. mmand were selected Figure 4 17 shows the Menu Editor provided in GUI for users to create their menu Accelerator also can be used for users who would like to use their keyboard either than point to desired choose menu The tag then will be attached to their callback to function certain command 54 T 4 RENT ERG ES jh ors Figure 4 17 Menu Editor The Property Inspector can be used to change fonts size label tag and so on in order to meet the best performance of the created program This can be shown by Figure 4 18 where the programmer can easily modify it Spe ay eer Laad 0 bore Dat Bashy rou red Cao Bei regDaaka d Buna ch CO ale Callb tk Cra auf Enab r Edeni Fanten ge Fanihisme Fond ints Ee S HE Pe ori rm norms 10 0 pao ire Figure 4 18 Property Inspector Finally the main interface for the program was introduced to connect the main page to the program that can perform the types of disturbance and location classifying 55 POWER QUALITY DISTURBANCE AND LOCATION DETECTOR USING ANN IMPLEMENTATION OF NEURAL NETWORK IN CLASSIFYING THE TYPE OF DISTURBANCE AND ITS LOCATION Figure 4 19 Main Page of the Program 4 2 3 Implementation of the Program For the users that use the program for the very f
67. mulation and analysis based on his her own data file There are some requirements and limitation on the data format 1 The number of data must be at least 800 as the program 15 set to perform the analysis of 800 data only Otherwise error will occur 2 The sampling frequency of the data has to be set as 1 kHz or at the step of 0 001s 3 The data should be in per unit p u form 4 The data must compile notepad with extension dat 11 Troubleshooting The list of troubleshooting for the program 1 as follows a If error message appeared Undefined function or variables or the program cannot be run please check if all the required files for the program are already installed inside the default MATLAB folder b If the program still cannot run even all the required file 1s installed then check the MATLAB version either it is match the MATLAB program version or not The program 15 build under MATLAB 7 0 0 19920 R14 C Before analyze on any new data samples make sure the data 1 correct form which is no text descriptions on the start of data samples files The data should be in values or error will occur during running d For incorrect number of data the error message Data Dimension Not Match will be displayed e The extracted feature might slightly different 1f compared to the reference 1f the sampling frequency step unit of the data 1s different with the required or the data 1s not in per unit p u form
68. ncy will result in high amplitude of the coefficients In the second level of decomposition half of the signal will be removed As a result the frequency band of the signal now 1 half of the original signal The process will continue and the level needed depends on the sampling frequency or the highest cutoff frequency set in the analysis The amplitude of coefficients increased as the level of decomposition increased until it reached certain level to start decreased The drop of coefficient s amplitude shows that the signal 15 completely decomposed 15 2 2 3 The Advantages of Wavelet Transform 1 The Wavelet Transform is a time frequency analysis which its coefficients contains information on successive bands or frequency Therefore it would be an advantage for the users who did not sure about the frequency that produced the signal which it 1s really well suited for disturbance signal with rapid changes 2 For comparison between Wavelet Transform and the popular Fourier Transform Wavelet provides different windows in order to isolate signal discontinuities such as if we would like to have some very short basis functions and on the same time too to have some very long basis functions which it will gives detailed frequency analysis Unlike Fourier Transform the resolution of the analysis 15 same for all frequencies as it needs to fit the window s width Frequency Figure 2 9 Fourier Basis Functions Frequency Fi
69. njadi rosak dan ia juga mampu menyebabkan kehilangan bekalan kuasa selama beberapa hari Oleh itu projek ini memperkenalkan cara terbaru untuk mengenal pasti lokasi dan jenis gangguan dengan lebih cepat dan mudah dengan menggunakan system pengesanan automatic yang menggunakan perisian Pendekatan yang digunakan dalam membangunkan projek ini ialah Discrete Wavelet Transform Analysis dan Artificial Neural Network Dari PSCAD litar akan direka berdasarkan jenis gangguan iaitu kenaikan voltan kejatuhan voltan pengsuisan kapasitor harmonic dan outage Hasil simulasi akan menunjukkan bahawa setiap jenis gangguan akan menghasilkan gelombang yang berlainan dari gelombang sinus yang asal dan oleh itu Wavelet Transform akan digunakan untuk menghasilkan ciri ciri gangguan Seterusnya ANN akan mengesan jenis gangguan dan lokasi di talian penghantaran Akhirnya program yang dinamakan Power Quality Disturbance and Location Detector Using ANN diperkenalkan di mana ianya melibatkan penggunaan Wavelet Toolbox Neural Network Toolbox dan GUI dari MATLAB CHAPTER TABLE OF CONTENTS TITLE THESIS STATUS SUPERVISOR S DECLARATION TITLE PAGE DECLARATION DEDICATION ACKNOWLEDGEMENT ABSTRACT ABSTRAK TABLE OF CONTENTS LIST OF FIGURES LIST OF APPENDICES INTRODUCTION 1 1 Project Background 1 2 Problem Statement 1 3 Project Objectives 1 4 Project Scope PAGE ii iii iv vi vii xiii Aa Wo N E Vil LITERATURE REVIEW
70. ntium 4 1 63 GHz processor higher speed recommended Operating system Windows 2000 XP Memory RAM 512 MB or higher Hard disc space 3 0 GB More space may be required to store data Other Peripheral A CD ROM or DVD ROM drive mouse and keyboard Software Development Hardware and Software Specifications The following 1s the environment used to develop and testing the software Category Specifications used Processor Intel Core Duo 1 83 GHz Processor Operating system Windows XP sp2 Memory RAM 512 MB DDR 2 Hard disc space 80 GB Software MATLAB 7 0 0 19920 R14 Other peripheral A DVD RAM drive a mouse and keyboard Installation of Program l Run MATLAB by clicking on the shortcut Make sure the MATLAB installed correctly 2 After MATLAB is ready click on the file and select under the New tab Click on Open Existing GUI and browse the file with name interface 1 Before that make sure the needed file already placed inside the MATLAV work folder See Ne 5 guai Fee Ferre amp A Bj B aw 6 irr 2 DER ee acra mee 4 A abd miue HH JM mol PAZI Bja SO VET Hed ng ES m 1 ERE GREEN T s w
71. o Property Inspector Object Browser Figure Activator Ido 1 Eie LeyOm TOO 12 See E Eu possem Se att b untitled fig ET Buton FR Toggle Button Radia Sutton Checkbox Component zm Edt et Palette X Static Test Slider L Freme Listhor Poop Menu in Axes E Figure Resize Tab Figure 4 14 GUI Layout Editor 22 The process of implementing a GUI involves two basic tasks which are laying out the GUI components and programming the GUI components First of all component that were selected to be use in creating Power Quality Disturbance and Location Detector using ANN were arranged inside the layout area The components used for this project are push button static text listbox axes and panel Figure 4 15 shows the layout of the program and Figure 4 16 shows the activated figure for 1t Figure 4 15 Layout of the Program 53 E 51 DU rm Figure 4 16 The Activated Figure for the Program The M Files were connected with the created GUI by using callback Every component has its own callback and we can assign different M File to be generated when different push button was selected In order to make it more like efficient file menu and help menu were added to the program which contains different function when different co
72. oe P dini IR EN ZEN me TS THIES BN Ji mem LP m BE ck rim L 1185 FER DD SW SIRE fit i L a px 155 FE Tilt 1 Pin FB a muB Pl amp zt ms imd Tr WIZL EN Wich aE 16787 i EE minm E d EN i zx After select the file interface 1 click on open and wait till GUI finished initialized aia pe n D ak iA s NN P nensis ome 88 rachel mum nom Meum cam ewe 8 04 waCDPIR BAILA keii 17 ee m aima mH 2 XE x IET mampi soit Ba t dro rem eS TUE ET lim T Ela gil Ap I Lire rir cua d p VI ri inia gem el i pli wap d inta EE PE mah iach a mud Drs BN gulis EE Bd i AED ER dl 1 5 After a GUI window appears click on Run
73. ork provides good technique for interpolation in dimensional space and compared to the Multi Layer Perceptrons MLP it is better where it can be trained using 2 fast stages of training algorithm without the need for time consuming non linear optimization techniques 27 2 3 5 The Advantages of Artificial Neural Network 1 Non linearity a Neuron is a nonlinear device b With a proper trained of neural network it will performed highly non linear mapping 2 Learning a The neural network is learn from the interaction with environment oy Generalization capability a Generalize the training information to similar situations which never experienced before 4 Complex mapping a Can synthesize complex mappings which maybe difficult or impossible 1f in mathematical form 5 Robust and fault tolerance a if the input data is incomplete or noisy the ANN still can produce satisfactory result b due to the distribution of computational load across many processing units the network gained same degree of fault tolerance with respect to processor failures 6 High speed a Can solve the mapping problem much faster than conventional methods and other intelligent methods such as expert system 23 CHAPTER 3 METHODOLOGY Generally the methodology involved for this project can be divided into gathering information on modeling circuit for disturbance and then simulating it to collect the data extracting characteristics for each types of distur
74. ponding wavelet transformed signal This can be mathematically expressed as below N N J N Y foy Y Y Y Equation 3 1 n l n l j l n l where f n Time domain disturbance signal in study nf N Total number of samples of the signal N 2 f Total energy of the f n signal n l N gt n Total energy concentrated the level i of the approximated version of n l the signa N Dy Total energy concentrated on the detail version of the signal from level n l Ms J I ep to T 34 Wavelet transtorm decomposition Energy curve 15 plotted mM T Fundamental Compare Energy deviation curve unique eatures patterris Figure 3 6 The Approach of Features Extraction to Obtain the Unique Pattern of Disturbance Step 4 In this stage the steps 1 2 and 3 are repeated for the corresponding pure sinusoidal version of the signal in study Step 5 total distorted signal energy of the signal in study found in step 3 15 compared to the corresponding one of the pure signal version evaluated in step 4 The result of this comparison 1s a deviation the will be the unique feature or pattern of the disturbance 35 ed i SEO ID 100 Equation 3 2 eref 4 where i Wavelet transform decomposition level ed i Vo Deviation between the energy distributions of the disturbance signal in study and its correspond
75. r it 1s favorable or not the target for each input must be achieved so that the type of disturbance and location that will be display is able to show the ability of neural network to classify the given input wisely Figure 4 41 shows the network architecture for probabilistic neural network that designed before It clearly shows that two layer was used which are iput weights layer and second layer weights that are set to the matrix of target vectors It was generated by MATLAB during running the file in command window net 73 Heural Network object architecture numinpurs numLayers 2 HiasConnect 1 0 inpurcConnectr layerConnect n autputConnect targetConnect 5 read only numTarcdgets read only numlnputDbelays read onls numLayerLbelavys read only subobject structures inputs 1 1 cell of inputs layers 2 1 cell of layers outputs 1 2 cell containing 1 output targets 1x2 cell containing no targets biases 221 cell containing bias inputWeights 2 1 cell containing input weight layerWeights 2x2 cell containing layer weight functions adaptFen initFen performFcen rtrainFcen parameters adaptParam initParam performParam trainParam none none none none none none none weight and bias values IW LW b other userdata 1 cell containing 1 inpu
76. rk Toolbox also provides two alternatives for users either to access it using command line or graphical interactive tools that are provided within MATLAB Both of it can be used to create train and testing the neural network For neural network the command line approach also was chosen for this project 28 GUI Builder GUIDE Alignment Tool Menu Editor Property Inspector Object Drowser Figure Activator Unclo d E uniitled fig File view Ss Gr id Button Toggle Autan F adio Button def Checkbon wp Edil Teal _ gt Coriponeat Palatte 5 a Slider Frame l ListEox Ea Menu Figire Resize Tab Figure 3 3 Layout Editor of GUI Builder GUIDE the MATLAB Graphical User Interface Development Environment provides a compilation of tools for creating graphical user interfaces GUIs These tools greatly help to simplify the process of designing and building GUIs Users can use the GUIDE tools to l Lay out the GUI Using the GUIDE Layout Editor 2 Program the GUI 29 By using the GUI builder users can create layout for their program easily by clicking and dragging GUI components such as panels buttons text fields sliders menus and so on into the layout area To program the GUI GUIDE automatically generates an M file that 15 able to controls how the GUI operates The M f
77. s similar to the surge arrester and surge diverter When the surge 15 detected 1t provides short circuit which will send the surge flow to the earth between the power line and ground 11 2 2 Wavelet Transform Wavelet transform is a transformation to basis function which 15 represented scale and time It transforms a waveform or a signal into its frequency component and this gives a faster and effective way to analyze non stationary voltage and current waveforms Wavelet transform can be classified into Continuous Wavelet Transform CWT and Discrete Wavelet Transform DWT 2 2 1 Continuous Wavelet Transform Continuous Wavelet Transform transforms a signal into wavelets that has small oscillations and highly localized in time It produces a time frequency with a good time and frequency localization CWT can be defined as f n n v s Six yy Equation 2 1 The equation 2 1 shows that the convolution 1s used directly as CWT 15 a convolution of data sequence with a scaled and translated version of the mother wavelet the psi function while the equation 2 2 15 alternative way by using the FFT based fast convolution 12 lt 21 5 100 uf W s FFT E jx 50 Je Equation 2 2 where E nip A A 8 k H Equation 2 3 N and N 2 2nk make Am 2 Not Not Equation 2 4 with x discrete data series N data series length s wavelet scale n localiz
78. s to vectors while vec2ind 1 vice versa It allows indices to either be represented by themselves or as vectors containing a in the row of the index they represent In order to display types of disturbance and the location of where it occurred basic structural programming language of MATLAB was used For example if the data from the wavelet extraction fulfills the requirement needed and match any of the target that already specified pnn it will give to the users the results by display the type of disturbance and location of it 4 2 2 GUI Design After all the scripts were written in M File then GUI will be used to compile all of them for easy use to the users For the first time designing GUI as we choose new GUI from the file menu at MATLAB figure as will appeared Users can choose the GUIDE templates they would like to start with and for this project the Blank GUI default was chose 51 GUIDE Quick Start Create Mew GIJI _ Existing GUI iSUIDE templates Preview Blank GUI Default ad GUI with Uicontrals dl GUI with Aves and Menu ET Modal Question Dialog Save on startup as C MATLAB warkiuntitled fid Figure 4 13 GUIDE Quick Start For the blank GUI component palette will be rearranged in order to perform a new layout of the created program The component palette will be put inside the layout area which can be size adjusted Algnment Tool Menu Ecit
79. stances away from the fault 45 4 1 3 Outage Simulation Outage occurred as there is zero voltage temporarily happened For the outage the breaker once again is placed in the diagram but now it is placed at the fixed load and the voltage also was measured at load Source Transformer Transmission Transformer Source Line 1 Transmission Line Transmission Line 2 3 Transformer Transformer Breaker Load Voltage Test Fixed Load Fixed Load Figure 4 7 The Schematic Diagram for Outage Disturbance Point 46 The breaker was set to operate at 0 2 second for 0 05 seconds which means it close at 0 25 second The breaker was placed at the load to produce the effect of outage and the voltage output was measured at the load The voltage supplied to the load will be temporarily affected which it will reduced to zero for that period The outage simulation was run for 0 8 seconds which produced the voltage output as shown in Figure 4 8 Main Graphs Va See SS 0 00 0 10 0 20 0 30 0 40 0 50 0 60 0 70 0 80 Figure 4 8 Sample of Output Voltage for Outage The sample of data 1s as shown in Figure 4 9 The first column indicates the time while the remaining column shows the voltage for phase A B and C in kV The data was saved in dat for easier modification and use in MATLAB The simulation was repeated by changing the length of transmission line 2 in order to collect data at different
80. t weight matrix Cell containing 1 layer weight matrix 1 cell containing 1 bias vector user stuff Figure 4 41 The Probabilistic Neural Network Architecture Probabilistic neural network was chose as it 1s simpler compared to feedforward neural network Besides that it can produce accurate output which the output produced will definitely be the same as target while for feedforward neural network the output will be only nearly to the target because it depends on the number of epochs and learning rate 74 CHAPTER 5 CONCLUSIONS AND RECOMMENDATIONS 5 1 Conclusions Power quality problems required a new and fresh idea in order to recognize the types of disturbance and location in a faster and simpler way so that the system can be restore back as soon as possible to avoid more severe damage and reduce the cost of it Therefore the goal of this project 15 to introduce a new way that involved the implementation of neural network to simplify the automated recognition system for the types of disturbance that occurred and it location The methods used in order to reach the goals are wavelet transform and neural network Wavelet transform 1s used to create characteristics for each types of disturbance while neural network then will be applied to classify the types of disturbance and also the location of where it occurred Finally the Power Quality Disturbance and Location Detector was successfully developed at th
81. to function This simulation program is developed to perform the classification of selected power quality types It will also perform the detection for location of disturbance occurred It is Wavelet Analysis based Neural Network simulation program Wavelet Transform Discrete Wavelet Transform is being used to decompose and extract the signal based on the frequency band that consists the signal Neural Network then 15 applied order to categorize types of disturbance and location based on the Wavelet Extraction result The approach that used to obtain the extracted feature with Wavelet Analysis 1s based on the Parseval s energy theorem while Deubechies was chosen as type of wavelet The extracted feature 1s the deviation of disturbance signal and pure sinusoidal signal which is as reference signal This data is then fed into Probabilistic Neural Network to identify the disturbance occurred and its location The simulation program provides an environment to visualize the extracted feature for power quality disturbance using Wavelet Analysis and also the recognition which shows the detected types of disturbance and the location Installation and Setup Information Recommended Hardware and Software Requirements The minimum requirement for the program to be able to perform 1s as same as the Matlab installation minimum requirements To run the program sufficiently it is recommended to use the specifications as below Processor Pe
82. uences for the long term problem of swells It usually occurred when transformer supply tap settings are set incorrectly and loads have been reduced It can cause overheating unnecessary tripping of downstream circuit breakers and putting stress on equipments e Harmonic Figure 2 5 Harmonic Harmonics can be categorized under waveform distortion There are five primary types of waveform distortion which are DC offset harmonics interharmonics notching and noise Harmonics can be defined as a distortion of waves from a pure sinusoidal at multiples of fundamental frequencies The contributors to harmonics are high power types of equipments using phase angle control and uncontrolled rectification and saturation of transformers core during energization with the increasing use of FACTS from power utility equipment It can give effects on transformers heating and system halts f Momentary Interruptions Figure 2 6 Momentary Interruptions Momentary interruption is a state where the voltage will be zero for 30 cycles to 2 minutes It 1s almost like a temporary blackout which can be experienced when the power supply suddenly did not operate and the common causes for this problem are equipment failure circuit breaker tripping and damaged of electricity supply grid When someone use computer and suddenly this interruptions happened at that time it will cause computers to lose data There are four types of interruptions which are instant
83. under IEEE 1159 the disturbances can be categorized under seven categories and for more accurate categorization each category will have their sub categories which depends on amplitude frequency spectrum modulation and others to identify them Therefore studies on type of disturbance and the software that are going to use 16 needed for this project For this project only certain types will implemented which are a Sag b Swell c Harmonic d Outages e Capacitor Switching This project also covered on circuit designation and simulation using PSCAD and it involved the usage of Discrete Wavelet Transform which 1 a type of wavelet transform to extract the characteristics of each disturbance based on the waveforms produced via simulation After that Probabilistic Neural Network will be applied to identify the types of disturbance and its location on the transmission lines CHAPTER 2 LITERATURE REVIEW 2 1 Power Quality Power quality is defined as any power problem occur involving voltage current or frequency deviation that results failure or misoperation of customer equipment while disturbance is temporary deviation from steady state waveform Therefore power quality disturbance can be described as a phenomenon which can cause the output voltage to be not in purely sinusoidal It 1s easily to react with sensitive equipments which can cause it to malfunction and as the power system 1 interconnected to each other
84. well Overvoltage Harmonic Memontary Interruptions Sub band Filtering of DWT Wavelet Decomposition Process Fourier Basis Functions Deubechies Wavelet Basis Functions Neuron An Artificial Neuron An Artificial Neural Network Threshold Feedforward Propagation Feedback Propagation Radial Basis Function Network The PSCAD Software The MATLAB Software The Layout Editor of GUI Builder Script Writing Wavelet Decomposition Tree PAGE N N WD 3 6 3 7 3 7 4 1 4 2 4 3 4 4 4 5 4 6 4 7 4 8 4 9 4 10 4 11 4 12 4 13 4 14 4 15 4 16 4 17 4 18 4 19 4 20 4 21 4 22 4 23 4 24 4 25 4 26 The Approach of Features Extraction to Obtain the Unique Pattern of Disturbance Probabilistic Neural Network Flow Chart Process Schematic Diagram for Voltage Sag Disturbance sample of Output Voltage for Voltage Sag sample Data of Voltage Sag Schematic Diagram for Voltage Swell Disturbance sample of Output Voltage for Voltage Swell sample Data of Voltage Swell Schematic Diagram for Outage Disturbance Sample of Output Voltage for Outage Sag Sample Data of Outage Schematic Diagram for Harmonic Disturbance Sample of Output Voltage for Harmonic Sample Data of Harmonic GUIDE Quick Start GUI Layout Editor Layout of the Program Activated Figure for the Program Menu Editor Property Inspector Main Page of Power Quality Disturbance and Location Detector Using ANN The Main Page of the Program The Tutori
85. wn as Figure 4 1 Source Transformer Transmission Transformer Source Line 1 Timed Fault Fault Location Transmission Line Transmission Line 2 3 Fixed Load Fixed Load Figure 4 1 Schematic Diagram for Voltage Sag Disturbance Load Voltage Test Point The fault was applied at the transmission line 2 for a duration 0 05 seconds and start after 0 2 second The applied fault causes the voltage drop and also causes most of the source current flow into the line The waveform produced by this circuit run for 0 5 seconds with plot steps of 0 001 s is shown in Figure 4 2 42 Main Graphs WV 0 00 0 10 0 20 0 30 0 40 0 50 Figure 4 2 Sample of Output Voltage for Voltage Sag The graph shows that the voltage drop at 0 2 second for duration of 0 05 seconds and then the voltage was restored back to normal The output data from PSCad contained one time value and load voltage for phase A C This information was then placed in separate columns and rows for each sample 0000000000000 0000000000000 0000000000600 0000000000000 10000000000000 2 13347236731568 b5 8952202309670 45604965578102 Z0000000000000E 02 85511145167430 1 4865670194143 53145556774003 s0000000000000E 02 367621305263 2 21015 49412142 17339281068734 40000000000000E 02 3409775217166 2 32
86. x and vector formulations 1n a fraction of the time 1t would take to write a program in a scalar noninteractive language such as C or Fortran a Wavelet Toolbox The Wavelet Toolbox is a collection of functions built within the MATLAB Technical Computing Environment It offered tools for the analysis and synthesis of signals and images and tools for statistical applications using wavelets and wavelet packets within the framework of MATLAB The Wavelets Toolbox provides two categories of tools l Command line functions 2 Graphical interactive tools 23 The first category of tools is made up of functions that users can call straight away from the command line or from their own applications Most of these functions are M files sets of statements that implement specialized wavelet analysis or synthesis algorithms The second category of tools 1s a type of graphical interface tools that afford access to extensive functionality For this project the first category was chosen to perform the Wavelet analysis which consists of Wavelet decomposition determination of the Wavelet coefficients and Wavelet extraction b Neural Network Toolbox Neural Network Toolbox 1s a set of established procedures that are known to work well It 15 a useful tool for industry education and research which will help users find what works and what doesn t and a tool that will help develop and extend the field of neural networks The Neural Netwo
87. ystems 24 25 PSCAD completely provide users with a library of pre programmed and tested models which includes from simple passive elements and control functions to more complex models such as electric machines FACTS devices transmission lines and cables If a certain desired model does not exist PSCAD provides the flexibility of building custom models either by assembling it graphically using existing models or by utilizing an intuitively designed Design Editor 3 1 2 MATLAB fhe Language of fechnical Computing Version 7 0 0 19920 R14 May 06 2004 License Number 198734 uTm Copyright 1384 2004 The MathWorks Unc Figure 3 2 MATLAB Software 26 MATLAB stands for matrix laboratory It is a high performance language for technical computing which combines computation visualization and programming in an easy to use environment where problems and solutions are expressed in familiar mathematical notation Typical uses include a Math and computation b Algorithm development c Data acquisition d Modeling simulation and prototyping e Data analysis exploration and visualization f Scientific and engineering graphics g Application development including graphical user interface building MATLAB is a system that does not require dimensioning as the basic data element is an array and this allows users to solve many technical computing problems especially those with matri

Untitled

Contents

Download Pdf Manuals

Related Search

Related Contents