CHAPTER 2 2 PROPOSED DIFFERENTIAL EVOLUTION BASED IDWNN CONTROLLER FOR FAULT RIDE-THROUGH OF GRID CONNECTED DFIG

Transcription

1 26 CHAPTER 2 2 PROPOSED DIFFERENTIAL EVOLUTION BASED IDWNN CONTROLLER FOR FAULT RIDE-THROUGH OF GRID CONNECTED DFIG 2.1 INTRODUCTION The key objectve of wnd turbne development s to ensure that output power s contnuously ncreased. It s authentcated that WT system supply the necessary reactve power to the grd at the tme of fault and after fault to ad the flowng grd voltage. At ths juncture, ths work ntroduces a novel heurstc based controller module employng Dfferental Evoluton and Neural Network archtecture to mprove the LVRT rate of grd connected wnd turbnes, whch are connected along wth DFIGs. The tradtonal crowbar-based systems were bascally appled to secure the RSC durng the occurrence of grd faults. Ths tradtonal controller s found not to satsfy the desred requrement, snce DFIG durng the connecton of crowbar acts lke a squrrel cage module and absorbs the reactve power from the grd. Ths lmtaton s taken care n ths thess by ntroducng heurstc controllers that remove the usage of crowbar and ensure that wnd turbne system supply necessary reactve power to the grd durng faults. The controller s desgned n ths chapter to enhance the DFIG converter durng the grd fault and ths controller takes care of the rde-through fault wthout employng any other hardware modules. Ths work ntroduces a

2 27 Double Wavelet Neural Network Controller whch s approprately tuned employng Dfferental Evoluton. To valdate the proposed controller module, a case study of wnd farm wth 1.5 MW wnd turbnes connected to a 25 kv dstrbuton system exportng power to a 12 kv grd through a 3 km 25 kv feeder s carred out by smulaton. 2.2 PROBLEM FORMULATION For the past few years, DFIG occuped the world s largest share of wnd turbne system as a varant to tradtonal varable speed generators. The desgned system should be n a poston to operate over a wde range of wnd speed for achevng optmal effcency n order to follow the optmal tpspeed value. Hence, the generator s rotor s to be desgned to operate n a varable rotatonal speed. DFIGs are therefore desgned to operate under both super and sub synchronous modes wth a speed range of the rotor to be n accordance wth that of the synchronous speed. The stator crcut module s connected drectly to the grd; on the other hand the rotor wndngs are connected to a three-phase converter. In case of a varable speed system, when the speed range requrements are low the DFIG provdes satsfactory performance and ths s found to be enough for the speed range employed wth the wnd power. In DFIG module, an Alternatng Current (AC)-to-AC converter s ncorporated n the nducton generator rotor crcut and the converter s to be rated to handle a fracton of the total power. The rotor power s around 32% of the nomnal generator power. As a result, the occurrences of losses n the converter module can be reduced n comparson wth that of the system where the converter s to handle the complete power, and the entre cost s mnmal because of the partal ratng of the power electronc crcuts.

3 28 Chapter 1 presented a detaled lterature study on the applcablty and desgn of DFIG for wnd turbnes. The control strateges are complex n nature yeldng to unstable system and also the voltage ratng of rotor s hgh enough n comparson wth that of the Mega Watt ratng of the wnd turbne generators whch possess stator to rotor wndng turns rato between 2.5 to 3. In ths work, a novel control strategy employng DE based IDWNN controller s proposed whch makes the DFIG based system hghly stable and the desgned system smulaton s carred out choosng approprate turns rato. Ths chapter also attempts to handle the FRT capablty of wnd turbne system and to make them more strngent. 2.3 MODELING OF DFIG WIND TURBINE SYSTEM n Fgure2.1 The schematc dagram of the wnd turbne and DFIG are shown Fgure 2.1 Wnd Turbne and the Doubly-Fed Inducton Generator system (Source: ( /meda/ mage3.png))

4 29 The AC-DC-AC converter system of DFIG s found to be dvded nto two components: the Rotor Sde Converter (C rotor ) and the Grd-Sde Converter (C grd ). Both C rotor and C grd are Voltage-Sourced Converters whch employ forced commutated power electronc devce lke Insulated Gate Bpolar Transstors (IGBTs) for syntheszng an AC voltage from a DC voltage source. On the DC sde, a capactor s connected whch acts as a DC voltage source. For connectng C grd to the grd, a couplng nductor L s been used. Slp rngs and brushes are used to connect the three- phase rotor wndng to C rotor and the three phase stator wndng s drectly connected to the grd. The power acqured by the wnd turbne s converted nto electrcal power by the nducton generator and ths s transmtted to the grd by the stator and the rotor wndngs. The control system module s employed to generate the ptch angle command and the voltage command sgnals rotor voltage, V r and grd voltage, V gc for C rotor and C grd respectvely. These sgnals tend to control the power of the wnd turbne, the DC bus voltage and the voltage or reactve power at the grd termnals. The equatons employed for computng mechancal power and the stator electrc power output are as follows: P m = T m r (2.1) P s = T e ω s (2.2) Where, P m - Mechancal power captured by the wnd turbne and transmtted to the rotor P s - Stator electrcal power output T m - Mechancal torque appled to rotor T e - Electromagnetc torque appled to the rotor by the generator ω r and ω s are the rotatonal speed of rotor and of the magnetc flux n the argap of the generator respectvely.

5 3 In case of lossless generator the mechancal equaton s gven by: J dωrdt = T m T e (2.3) Consderng steady state at constant speed, n case of lossless generator T m = T e and P m = P s + P r, where P r s the rotor electrcal power output and J s the rotor and wnd turbne nerta coeffcent. The rotor electrcal power output s gven by, P r = P m P s = T m ω r T e ω s = T mωs ωrωs ω s = st m ω s = sp s (2.4) where s s defned as the slp of the generator: s = (ω s ω r )/ω s. The turbne and trackng characterstcs s as gven n Fgure 2.2. ( Fgure 2.2 Turbne and trackng characterstcs

6 31 In ths case, power s controlled such that t follows a set powerspeed characterstcs called as trackng characterstcs. The measurement of actual turbne speed s done and ts respectve mechancal power s consdered as the reference power for the power control loop. In fgure 2.2, ABCD curve represents the trackng characterstcs - reference power s zero from zero to A poston, the power between A and B s a straght lne and the speed of B s to be greater than that of A. In between B and C, the characterstc s the locus of the maxmum turbne power. Further, the characterstc s a straght lne from C to D DFIG Modelng (Theodoros et al. 214) DFIG modellng s carred wth an nducton machne controlled n a synchronously rotatng dq-axs frame, wth the d-axs found to be orented along the stator-flux vector poston. Ths approach forms Stator-Flux Orentaton (SFO) control and s used for modellng DFIG. Followng ths manner, a decoupled control exstng between the electrcal torque and the rotor exctaton current s noted. As a result, the actve and reactve powers are controlled ndependently from each other. Based on ths the stator and rotor voltage equatons are gven by, v v v ds qs dr R s ds R R s qs r dr d sqs dt ds d sds dt s s qr qs d dt dr (2.5) (2.6) (2.7)

7 32 v qr R r qr s s dr dqr dt (2.8) In the above equatons stator voltages and stator currents are gven by V s and I s, rotor voltages and currents are represented as V r and I r, stator and rotor resstances are gven by R s and R r, s and r are respectvely stator and rotor flux lnkage components. As the machne s rotatng n dq-axs frame, the ndex d represents drect axs component of the reference frame and q represent quadrature axs components of the reference frame. In each of the equatons (2.5) to (2.8), the gven flux lnkages are defned by, L L (2.9) ds qs s ds s qs m dr L L (2.1) dr r dr m qr L L (2.11) qr r qr m ds L L (2.12) m qs Where, L n each of the cases represent ther nductances and L m s the magnetzng nductance. The equatons (2.5) through (2.12) represent the modelled equatons for the consdered DFIG system Descrpton of the Conventonal DFIG Control system Numerous research artcles (Govndarajan & Raghavan 214, Kong et al. 214, Wang, T et al. 214, WU et al. 214, Zhang et al. 215, Theodoros et al. 214) have presented the conventonal DFIG control system, whch possess two control modules: RSC Control system and GSC Control system. Ths secton presents the descrpton of the conventonal DFIG control system wheren n ths module no steps are taken to handle the FRT of the DFIG. Ths conventonal control module wth RSC and GSC wll generally be merged wth that of the hardware lke STATCOM or Crowbar system wth the lmtatons as dscussed n the lterature study of chapter 1. The conventonal

8 33 DFIG control system s presented here n order to make the lucd dfference between the applcablty of the proposed optmzed control module from that of the conventonal control system module Rotor Sde Converter Fgure 2.3 depcts the rotor sde converter control system. The reactve current flowng n the C rotor controls the voltage or the reactve power at the grd termnals. The man objectve of RSC s to regulate the stator actve and reactve power ndependently. To have decoupled control of the stator actve power, P s and reactve power, Q s, the rotor current gets transformed to d-q components employng the reference frame orented wth that of the stator flux. The quadrature axs current component, I qr controls the stator actve power and the reference value of the actve power, P ref as n Fgure 2.3 ( s obtaned usng the trackng characterstcs va a Maxmum Power Pont trackng. The measured stator actve power wll get subtracted from the reference value of the actve power and the error s drven to the power regulator. The output of the power regulator s the I qr_ref, whch s the reference value of the quadrature axs rotor current. Ths value wll be compared wth that of the actual value, I qr and that error s flowed through the current regulator, whose output s v qr, reference voltage for the quadrature axs component.

9 34 Fgure 2.3 Rotor Sde Converter Untl the reactve current s wthn the maxmum current values gven by the converter ratng, the voltage gets regulated at the reference voltage. On operatng the wnd turbne n Volt Ampere Reactve (VAR) regulaton mode, the grd termnals reactve power s mantaned constant by a VAR regulator. In ths work, a DFIG whch feeds a weak alternatng current AC grd s consdered, henceforth AC voltage regulator component s used and the reactve controller, VAR regulator module s not used. The output of the AC voltage regulator s the reference drect axs current I dr_ref, whch must be njected n the rotor by the rotor sde converter. The same current regulator as that of the power regulator s used to regulate the actual I dr component of postve-sequence current to ts reference value. The output from ths regulator s the drect axs voltage V dr, generated by rotor sde converter. The current regulator s provded wth the feed forward terms that predct V dr. V dr and V qr denote the drect axs and quadrature axs voltages respectvely. The magntude of the reference rotor current I r_ref s gven by,

10 35 I r_ref = I dr_ref + I qr_ref (2.13) The maxmum value of ths rotor current s lmted to 1 p.u. On the other hand when I dr_ref and I qr_ref are such that ther magntude s hgher than 1 p.u, then the I qr_ref component s reduced to brng the magntude to 1 p.u Grd Sde Converter Fgure 2.4 ( Fed-Inducton- Generator-n- wnd-power- generaton/ Image_5. jpg) shows the schematc of grd sde converter control system. The GSC control system s desgned to regulate the voltage of the DC bus capactor.e., to mantan the DC-lnk voltage constant. Also, GSC module s used to generate or absorb reactve power. Fgure 2.4 Grd Sde Converter GSC Control systems consst of a measurement system to measure the d and q components of AC postve sequence currents that s to be controlled as well as the DC voltage, V dc. The outer regulaton loop conssts of a DC voltage regulator. DC voltage regulator output s gong to be the reference current, I dgc_ref to be fed for the current regulator. The current n phase, I dgc wth that of the grd voltage that controls actve power flow s also measured and the dfference between I dgc and I dgc_ref s one of the nputs to the

11 36 current regulator. A current regulator forms the nner current regulaton loop and ths regulator controls the magntude and phase of the voltage generated by C grd from that of the I dgc_ref developed by the DC voltage regulator and gven I q_ref. It should be noted that the magntude of the reference grd converter current s gven by, I gc_ref = I dgc_ref + I qr_ref (2.14) The maxmum value of ths I gc_ref s lmted to a value defned by the converter maxmum power at nomnal voltage. When I dgc_ref and I q_ref are such that the magntude s hgher than ths converter maxmum power, then the I q_ref component s mnmzed to revert back the magntude to the maxmum value. The DC voltage s controlled wth the sgnal I dgc_ref and the reactve power s controlled by means of I q_ref from the reactve power regulator (Ca et al. 213, Lu, H et al. 214, Mohsen & Islam 212). 2.4 DESCRIPTION OF THE PROPOSED DE-IDWNN OPTIMIZED CONTROL SYSTEM The conventonal DFIG control system dscussed n Secton 2.3 does not take nto account the fault rde-through of the nducton generator. Thus n ths secton, attempt s made to carry out rde-through fault elmnatng the extra hardware component. The optmzed control module s desgned n a manner to accomplsh optmal synchronzaton between the RSC control system and GSC control system and as well can handle the dsturbances occurrng n the system because of the fault. Ths s taken care n the system even wth the wnd turbne feedng a weak AC grd. At ths juncture, t should be noted that the proposed controller s to perform effectvely wthn a short span of tme and as well has to be not nfluenced by the measured nose that

12 mght nterrupt n the system. Ths controller s also to be desgned to take care of lackng machne parameters nformaton of the system. 37 All the above dscussed condtons wll result n added nonlnearty of the system. Hence, to address the sad dffcultes as well to handle the unavodable non-lnearty ntroduced n the system, the proposed DE- IDWNN controller desgn lay on the evolutonary computatonal ntellgence technques and wll provde a better soluton n comparson wth that of the conventonal approaches. To be more accurate the controller desgn s performed wth a double wavelet neural network controller, whose weghts are optmzed and tuned employng dfferental evoluton. The controllers employed are IDWNN controllers and the tunng of IDWNN controller meets the fault rde-through. Further, the applcablty of DE also performs weght optmzaton of the NN controller to acheve better soluton and faster convergence Dfferental Evoluton Algorthm A populaton based stochastc evolutonary algorthm ntroduced by Storn & Prce 1995, s Dfferental Evoluton algorthm whch s n a way smlar to genetc algorthms, Storn & Prce 1997, employng the operators crossover, mutaton and selecton and searches the soluton space based on the weghted dfference between the two populaton vectors. To dfferentate: genetc algorthmc approach reles on crossover and dfferental evoluton approach reles on mutaton operaton. Mutaton operaton s used n DE algorthm as a search mechansm and the selecton operaton drects the search towards the prospectve regons of the search space. Intally n DE, populatons of soluton vectors are generated randomly at the begnnng and ths ntally generated populaton s mproved over generatons usng mutaton, crossover and selecton operators, Prce et al. 26. In the progress

13 38 of DE algorthm, each of the new soluton resulted competes wth that of the mutant vector and the better ones n the race wns the competton. The standard DE s gven n the Table 2.1. Table 2.1 Pseudo code of standard DE Start Stop Intalze the populaton When (stoppng condton s false) perform the followng Begn Perform Mutaton Operaton Update the new soluton value Perform Crossover operaton Modfy the tral populaton vector Evaluate the ftness functon Perform Selecton operaton to select the best operator End Proposed Inerta based Double Wavelet Neural Network The NN archtecture of the proposed Inerta based IDWNN s shown n Fgure 2.5. ( fgure/fg5/archtecture-of-nerta-based-double-wavelet-neural-network_ small. png). It conssts of Input layer, Hdden Layer 1, Hdden Layer 2 and the Output Layer. The hdden layer 1 contans n-wavelet synapses wth h1 wavelet functons and the hdden layer 2 contans one wavelet synapse wth h 2 wavelet functons.

14 39 Fgure 2.5 Archtecture of Inerta based Double Wavelet Neural Network Vector sgnal; xk x k x k,..., x k 1, 2 n where; k,1,2,..., n whch s the number of sample nputs n the tranng set. The output of the proposed Inerta based Double Wavelet Neural Network s expressed as: y k f n 1 f x k f uk (2.15) j x k w j k w j (2.16) h 2 n h1 q q 1 j y h k uk w q k 2 q 2 (2.17) where; w j k : Synaptc Weghts The wavelets dffer between each other by dlaton, translaton and bas factors are realzed n each wavelet synapse. The archtecture of proposed IDWNN wth nonlnear wavelet synapses are shown n the Fgure 2.6.

15 4 Fgure 2.6 Archtecture of Proposed IDWNN wth non-lnear wavelet synapse The expresson for tunng of the output layer s gven by, where; E 1 1 (2.18) k dk yk e k d k : External tranng sgnal. The learnng algorthm of output layer of the proposed neural network s: k w k ke k uk w j 1 j j (2.19) The equaton (2.19) can be represented n vector form as; k w k k ek uk w 1 (2.2)

16 where; e k : Learnng Error, k Factor The analogy of the learnng algorthm s gven by, w 41 : Learnng rate parameter and : Inerta k w k k uk k e (2.21) w 1 w 1 2 k k u k w (2.22) 1 The range of s between and 1. The expresson for tunng of the hdden layer s: n h Ek dk f uk dk f 2 2 j x k w j k (2.23) 1 j Learnng algorthm of hdden layer of the proposed neural network are: w j ' k 1 w k k ek f uk x k (2.24) The equaton (2.24) can be represented n vector form as; w j ' k 1 w k ke kf uk x k where; e k : Learnng Error, k Factor (2.25) : Learnng rate parameter and : Inerta The analogy of the learnng algorthm n equaton (2.24) s gven by, w k w k w k uk x k w k ' e f 1 (2.26) w k 1 k x k 1 2 (2.27) The range of s between and 1. The propertes of ths proposed algorthm have both smoothng and approxmatng. The nerta parameter helps n ncreasng the speed of convergence of the system. The nerta parameter also prevents the network gettng converged to local mnma. The nerta factor s between values to 4. j 2

17 Proposed DE based IDWNN Optmzaton Controller Based on the gven nputs and outputs of the system, the proposed IDWNN model s desgned and the DE approach s employed to tune the weghts of the IDWNN and as well to tune the outputs of the NN model. The proposed pseudo code for DE based IDWNN optmzaton controller s as gven n Table 2.2. Table 2.2 Pseudo Code for DE IDWNN Controller Start Phase I: Intalze the IDWNN weghts, learnng rate and nerta factor. Randomly generate weght values. Present the nput samples to the network model. Compute net nput of the network consdered. Apply actvatons to compute output of calculated net nput. Perform the above process for hdden layer to hdden layer and hdden layer to output layer. Fnally compute the output from the output layer. Perform weght updaton tll stoppng condton met. Phase II: Present the computed outputs from IDWNN model nto DE. Invoke DE Perform Mutaton Perform Crossover Perform Selecton Do carry out generatons untl ftness crtera met. Note the ponts at whch best ftness s acheved. Present the ponts of best ftness back to IDWNN.

18 43 Phase III: Invoke IDWNN wth the nputs from the output of Phase II Tune the weghts to ths IDWNN employng DE Invoke DE Perform Phase I process for tuned weghts of DE. Contnue Phase II. Repeat untl stoppng condton met (Stoppng condtons can be number of teratons/ generatons or reachng the optmal ftness value) Stop The proposed DE-IDWNN controller s appled to handle FRT of grd connected DFIG and the methodology adopted s presented n the followng secton. 2.5 FAULT RIDE-THROUGH OF DFIG MODULE USING PROPOSED DE-IDWNN CONTROLLER DFIG module wth FRT employng proposed DE-IDWNN controller s as shown n Fgure 2.7. In the proposed desgn, GSC control system remans the same as that of the conventonal DFIG module and the modfcaton occurs n the RSC control system module.

19 44 Regular RSC Control System V dr Current Regulator V qr + Vqr dq abc PWM - Proposed optmzed controller r * N crf V dc * IDWNN FRT X V S_ref V s Fault Detecton Fgure 2.7 Proposed DE-IDWNN Optmzed Controller The proposed optmzed controller starts ts operaton only when the AC voltage V s exceeds n hgh of ten percent than that of the set reference voltage. The constrants that should be taken care for safe guardng the DFIG are the DC-lnk overvoltage and the rotor over current. Both these constrants should not exceed ther set lmts n the consdered restorng perod. Also, care should be taken n order to transfer the addtonal energy generated n the rotor va the converters to the grd. Ths process of transferrng the addtonal energy nduced wll enable the DC-lnk voltage and rotor current to mantan at ther respectve normal values. A setback on performng ths operaton s that, when the rotor current s decreased by fast transfer of the stored energy from rotor to grd, there mght be a chance that the DC-lnk voltage ncreases abruptly and t may get devated from the normal lmts. On the other hand, wthout fast mechansm, slowly decreasng the rotor current such that the DC-lnk voltage

20 does not exceed the lmt may result n the rotor current reachng abnormal transent ponts. Henceforth, for mplementng an effectve and effcent fault rde through, the transton sgnals of the rotor current should also consder the nomnal values of the DC-lnk voltage. The proposed controller s the nerta double wavelet neural network controller and the nput to IDWNN FRT s the V * dc and * r and the output of the Neural Network controller s the N crf. acts the nputs to the IDWNN optmzed controller and are gven by, 45 * V dc and * r V * dc V V dc ss (2.28) Vmax Vss (2.29) * r ss r max ss Where, V ss and ss ndcates the steady state values, V max and max specfes the maxmum values, r s the rotor rms current and V dc s the DC- lnk voltage. Both the nput values are normalzed before t s fed nto the Neural Network controller. Desgn of neural network controller s performed as shown n Table 2.3 and the tranng process of the controller s carred out to obtan the N crf. Orgnally the controller s desgned by consderng random weghts nto the IDWNN and then DE s adopted for tunng the weght values of the IDWNN controller. Non-lnear wavelet synapse s employed durng the tranng process for faster convergence. Table 2.3 Proposed IDWNN Controller parameters Parameters of the proposed IDWNN controller Set values Number of nput neurons 2 Number of output neurons 1 Inerta factor 1.75 Number of hdden layer neurons 8 Learnng rate parameter 1

21 46 As ndcated n secton 2.4.2, the nerta factor can take value from to 4. The proposed IDWNN module s smulated for varous range values between to 1 wth an nterval of.25 (.e.,,.25,.5,.75,...4). Durng the smulaton process, t s observed that when the nerta factor tends to 1.75 better soluton wth mnmal ftness s attaned. As a result, the nerta factor for the proposed IDWNN controller s set as Along wth the proposed IDWNN controller, dfferental evoluton s employed to tune the weghts of IDWNN controller as well to mnmze the objectve functon or the ftness functon as gven n equaton (2.3). 2 Vdcmax Vss r max ss mn f (2.3) Vmax Vss max ss 2 V max and rmax represents the maxmum value of the sgnals durng that of the entre restorng duraton. On applyng DE for the consdered problem, each set of varables are represented by bnary strngs called chromosomes and chromosomes are made up ndvdual enttes called genes. All the chromosomes generated result n the formaton of populaton. In ths problem under consderaton, populaton sze s 2.e, 2 chromosomes and the length of chromosome s taken to be 8 e., 8 genes comprse a chromosome. The ntal 6 genes represent the bnary strngs of the range of nput parameters and the remanng genes represent the output range. The process of DE s carred out as gven by the pseudo code n Table 2.1. DE s nvoked and actvated to search for mnmzng the optmzaton functon gven n equaton (2.3). In DE process, mutaton s carred out at the ntal process and then cross over and selecton operatons are performed. The search process s carred out to fnd the mnmzed value for the equaton (2.3) and the procedure s contnued untl a specfed stoppng condton s reached.

22 The mportance of the ftness functon n equaton (2.3) s as follows: Generally for a varety of problems, the objectve functon wll be an ntegral functon where the output s the complete behavour of the system n a partcular tme nterval. In ths problem doman, the am s to lmt the nstantaneous value of rotor current and DC- lnk voltage so as to elmnate the trppng of DFIG. As a result, the objectve functon s the sum of the squared components that are to be mnmzed. Also, t s to be noted n ths case, that the am here s not only to mnmze the specfed objectve functon n equaton (2.3), but also to mantan the balance between the rotor current and DC-lnk voltage n the range of acceptable values. Fgure 2.8 shows the three dmensonal graph for the IDWNN FRT output N crf wth respect to 47 * V dc and * r after that of the DE optmzaton. The proposed DE-IDWNN controller can be employed for varous szes of machnes. The tranng of the NN controller and the DE optmzaton process remans the same n all cases. Fgure 2.8 Three dmensonal output for the traned controller

23 SIMULATION RESULTS OF PROPOSED DE-IDWNN CONTROLLER The proposed DE-IDWNN controller s valdated by applyng the sad control strategy for a 1.5 MW wnd turbnes connected to a 25 kv dstrbuton system exportng power to a 12 kv grd through a 3 km 25 kv feeder. The electrcal system, Theodoros et al. 214, consdered for valdatng the proposed control desgn s gven n Fgure 2.9. Also, ths work bascally deals wth handlng three phase symmetrcal grd faults. The occurrence of three phase faults s noted at.5 seconds and the smulaton s carred out. The smulaton response s noted for both the tradtonal DFIG control system and as well that of the proposed controller. From the smulaton response of the tradtonal controller, t can be noted that ths system requres an auxlary system for fault rde-through opton and observes the lmtatons as dscussed n ntroducton secton. Durng the process of smulaton t s observed that the DFIG along wth the proposed controller handles the complete response at fault perods and after fault perods and gets rde-through fault elmnatng the applcablty of auxlary hardware. Table 2.4 shows the specfcatons for the parameters of the DFIG system and the grd system, Theodoros et al. 214, consdered for smulaton. The entre proposed control strategy s run n MATLABR29 envronment and executed n Intel Core2 Duo Processor wth 2.27GHz speed and 2. GB RAM. Smulnk envronment n MATLAB s used to model the converter modules. All the smulatons are carred out for wnd speed of 12 m/s.

24 49 Table 2.4 System Specfcatons of DFIG system and grd module Specfcatons of DFIG system Rated power 1.5 MW Rated Stator Voltage 69 V Nomnal Wnd Speed 12 m/s Stator Resstance.76 pu Rotor Resstance.5 pu Stator leakage nductance.1716 pu Rotor leakage nductance.156 pu Magnetzng nductance 2.9 pu Turns rato 2.7 Rotatonal nerta 5.4s Nomnal DC- lnk rated voltage 12 V Dc bus capactor 6 mll farad Specfcatons of AC Grd Rated Voltage 2kV Frequency 5 Hz Short crcut rato 2.23 Load 1 power 4 kw & 12 kvar Load 2 power 5 kw & 15 kvar Load 3 power 5 KW & 15 kvar Length of Transmsson Lnes Lne 1 and Lne 3 15 km Lne 2 and Lne 4 3 km Synchronous Machne Speed 15 rpm Synchronous Machne Power 85 kva

25 5 Fault PCC ac grd Lne 1 Lne 2 2kV/69V WT Load 1 Lne 3 Lne 4 Load 2 Load 3 2kV/4V Synchronous Generator Fgure 2.9 Electrcal system consdered for valdatng the proposed controller Fgure 2.9 Electrcal system consdered for valdatng the proposed controller Fgure 2.1 Response of three-phase stator voltage for tradtonal controller

26 51 Fgure 2.11 Response of DC-lnk voltage for tradtonal controller Fgure 2.12 Response of rotor current for tradtonal controller

27 52 Fgure 2.13 Response of stator current for tradtonal controller Fgure 2.14 Response of wnd turbne actve power output for tradtonal controller

28 53 Fgure 2.15 Response of wnd turbne reactve power output for tradtonal controller Fgure 2.16 Response of rotor speed for tradtonal controller

29 54 Fgure 2.17 Response of q-component of rotor voltage for tradtonal controller Fgure 2.1 Fgure 2.17 shows the response of the tradtonal controller generated for the wnd speed of 12 m/s. From Fgure 2.1 and Fgure 2.11, t can be observed that the DC voltage exceeds the set lmt resultng n damagng the capactor. Also, the rotor current ncreases n an advert manner, nearly 1% than that of the acceptable value n the rotor sde converter control system. Durng ths conventonal controller acton, t s noted that the entre response of the system vares n an erroneous manner resultng n more fluctuatons n the grd sde. Thus, the conventonal controller s modfed to enable approprate handlng of rde- through the fault condtons. The proposed DE-IDWNN controller s smulated for 1 generatons and the entre response observed durng the smulaton response are as shown through Fgure 2.18 through Fgure Non-lnear wavelet synapse s utlzed for the updaton process of IDWNN controller and the DE s ntated to tune the weghts of IDWNN controller and as well that of the objectve functon as gven n equaton (2.3). The evaluaton of the ftness

30 55 functon s studed at 85% voltage dp. The response of the observed parameters durng smulaton process proves the removal of fluctuatons occurrng durng the conventonal control desgn and as well noted to reach the steady state at the earlest. The Fault Rde-Through of the DFIG module s acheved at the pont wheren optmal soluton s arrved employng the proposed controller desgn. At ths juncture, over voltages noted at DC-lnk pont and over currents at the rotor sde are observed to be well wthn the lmt of the maxmum set threshold values. As a result, the damagng of the capactor s protected and fault and post fault occurrence to the grd are controlled n an effectve manner. Fgure 2.19 and Fgure 2.2 show the rotor current and stator current wth fluctuatons reduced and obtanng the steady state value at a faster rate. Table 2.5 Ftness functon evolved durng generatons Generatons Ftness functon (f) as n equaton (2.3) Fgure 2.21 and Fgure 2.22 show the response of the wnd turbne output actve and reactve power. Fundamentally, n ths proposed controller desgn, RSC wll not be cut durng the fault condton and ths RSC

31 56 provdes the requred reactve power to the grd, n case to handle the suffcent voltage drops occurrng. But the proposed DE-IDWNN controller acts to the fault wth an optmal soluton arrved and thus t wll prevent more amount of reactve power to be transferred from DFIG after the fault. In the proposed desgn, DFIG supples the grd wth the requred amount of reactve power and s found to sustan wth that of the grd voltage. At the tme of fault, the rotor speed of the proposed controller ncreases whch s seen n Fgure The ncrease n speed of the rotor s noted due to the capacty of the wnd turbne to store huge energy. At the occurrence of fault, there wll be a huge drop n the voltage and the wnd power s transferred as knetc energy to the rotor wthout dsspatng to that of the grd. At the end of the fault duraton, the grd receves ths energy and the ncrease n the speed of the rotor wll automatcally get settled to the earler value (.e., the value before the fault has occurred). Fgure 2.24 and Fgure 2.25 show the q-component and the modfed q-component of the rotor voltage of the proposed controller desgn. The occurrence of transents s noted and the proposed controller acts n a manner to brng back the ac voltage to ts set value resultng n the above sad transent. In case durng smulaton process, when there occurs a hgher voltage dp, DFIG s allowed to dsconnect. Ths s the bad condton on gettng connected to the grd. Thus the entre smulaton study s carred out for wnd speed at 12 m/s and voltage dp of 85%. The proposed controller s found to rde-through the fault occurred. Table 2.5 shows the ftness functon values evolved durng varous generatons. At the end of the 1 generatons, t s noted that the ftness value reached The DC lnk voltage and the rotor current are found to be mantaned wthn the sad permssble lmts at ths optmal pont.

32 Three Phase Stator Voltage (pu) FIGURE18(C) Tme (sec) Fgure 2.18 Response of three-phase stator voltage for proposed DE-IDWNN Controller Table 2.6 Comparson of the Ftness Functon values Methods employed Authors Optmal best Value of f GA Based approach Theodoros et al. (214) ANN Controller Dong et al. (211) ANFIS Controller Govndarajan & Raghavan (214) Proposed DE-IDWNN Controller 4.21 From Table 2.6, t can be observed that the proposed controller acheves a mnmal ftness functon value of 4.21 n comparson wth that of the earler other methods avalable n the lterature provng the effectveness of the controller. Fgure 2.26 show the varaton of ftness functon wth respect

33 Stator Current (kamps) Rotor Current (kamps) to generatons and from Fgure 2.26 t can be nferred that the search process enables the ftness functon to reach a mnmum value Tme (sec) Fgure 2.19 Response of rotor current for proposed DE-IDWNN Controller 2 FIG2(B) Tme (sec) Fgure 2.2 Response of stator current for proposed DE-IDWNN Controller

34 WT output recatve power (mvar) WT output actve power (mw) FIGUARE 21(A) Tme (sec) Fgure 2.21 Response of wnd turbne output actve power for proposed DE-IDWNN Controller Tme (sec) Fgure 2.22 Response of wnd turbne output reactve power for proposed DE-IDWNN Controller

35 6 Fgure 2.23 Response of rotor speed for proposed DE-IDWNN Controller Fgure 2.24 Response of q-component of rotor voltage for proposed DE- IDWNN Controller

36 Computed Ftness values Modfed q-component of the rotor voltage (pu) Tme (sec) Fgure 2.25 Response of the modfed q-component of rotor voltage for proposed DE-IDWNN Controller No. of Generatons Fgure 2.26 Varaton of ftness functon wth respect to number of generatons

37 SUMMARY Ths work proposed a novel heurstc based controller module employng dfferental evoluton and neural network archtecture to mprove the low-voltage rde through rate of grd connected wnd turbnes, whch are connected along wth doubly fed nducton generators. The lmtaton of the tradtonal control system s taken care n ths work by ntroducng heurstc controllers that removes the usage of crowbar and ensures that wnd turbne system supply necessary reactve power to the grd durng faults. The controller desgned n ths chapter enhances the DFIG converter durng the grd fault and ths controller takes care of the rde-through fault wthout employng any other hardware modules. The proposed controller desgn s valdated for a case study of wnd farm wth 1.5 MW wnd turbnes connected to a 25 kv dstrbuton system exportng power to a 12 kv grd. The results smulated prove the effectveness of the controller desgn n comparson to be better wth that of the methods avalable n the lterature. Even though the mutaton operated employed n dfferental evoluton s central to the success of ts search for weght values, they are noted to be computatonally expensve and consume more computaton tme. Further, the DE-IDWNN resulted n local optma durng the convergence when t s appled for the consdered problem. To overcome ths dffculty and to mprove the performance of the controller a hybrd PSO-GSO based Spkng Neural Controller s presented n the next chapter for carryng out the fault rde-through analyss n grd connected DFIG.