1992-8645 ARTIFICIAL NEURAL NETWORK BASED UNIFIED POWER .

Journal of Theoretical and Applied Information Technology15th September 2018. Vol.96. No 17 2005 – ongoing JATIT & LLSISSN: 1992-8645www.jatit.orgE-ISSN: 1817-3195ARTIFICIAL NEURAL NETWORK BASED UNIFIED POWERQUALITY CONDITIONER FOR POWER QUALITYIMPROVEMENTS OF DOUBLY FED INDUCTIONGENERATOR1kaoutar RABYI, 2Hassane MAHMOUDIElectrical Engineering Department, Mohammadia School of Engineers, University Mohammed V,Morocco1E-mail : rabyi.kaoutar@gmail.com1,2ABSTRACTTo reduce the mathematical operations and different transformations, the artificial neural network (ANN)approach is proposed for the unified power quality conditioner (UPQC). This paper proposes a voltagesource inverter (VSI) based UPQC with ANN controller when a Doubly Fed Induction Generator (DFIG) isconnected to the grid. The performance of UPQC with ANN controller is tested under different sag,harmonic and swell conditions, the algorithm used for the ANN control is Gradient Descent withMomentum to generate the referencing signals and maintain the UPQC dc link capacitor voltage. Thesimulations are carried out in the software Matlab/Simulink. Results shows efficiency of the ANN controlstrategy in compensating currents and voltages of the system.Keywords: DFIG, UPQC, ANN, Sag, Swell, Harmonic1.INTRODUCTION:In last few years, problems of powerquality are rising due to the grid codes of renewableenergy integration especially wind energy which isnot a predicted energy [1]. Variable speed windturbines pose many challenges concerning the faultRide through capability and the control of reactivepower [2]. The Doubly Fed Induction Generator(DFIG) is the most robust wind turbine due to itsvariable speed (fig. 1) , also its independence ofactive and reactive power control [3], even thoughthe DFIG is the most used type, it is very sensitiveto the electrical grid voltage interruption; during thefault, the generation of active power reduces due tothe drops of the voltage [4]; also, when integratinga large wind farms to the electrical grid, manyproblems occur like the voltage sag and swell,flicker and harmonics Fig. 1: schematic of the DFIGTo meet the requirements of sag and harmonicregulation, electronic power devices are used, theyprovide multiple function such as load balancing,power factor compensation, voltage flickerreduction and regulation, harmonic filtering, activeand reactive power control [5].When the grid side has a short circuit, the rotorcurrents increase, if there is no protection for theconverter opposing the high currents, the converterwill be deteriorated [6]. There are two operationmodes: compensation with series voltage usingDynamic Voltage Restorer (DVR) [7]; where thevoltage sags are not totally mitigated, so in thispaper, the proposed solution is a system using5977

Journal of Theoretical and Applied Information Technology15th September 2018. Vol.96. No 17 2005 – ongoing JATIT & LLSISSN: 1992-8645www.jatit.orgUnified Power Quality Conditioner (UPQC) for thevoltage compensation.UPQC is one of the important electronic device, itis used for the mitigation of harmonics current andvoltage sag [8], it is a single device of series andshunt active power compensators (fig. 2), the seriesconverter acts on the voltage grid, while the shuntconverter reduces the harmonics [9]. Thus, manyresearches have been done about the controlstrategies for the UPQC; the control system plays aE-ISSN: 1817-31952.DFIG MODELING:The rotor of the DFIG WT isconnected to the grid via slip rings via theRotor Side Converter (RSC) and the GridSide Converter (GSC), the stator isconnected directly to the gridfundamental role for power conditioner, a quickresponse of signal disturbance and rapid extractionof referencing signal are the major requirement fora good compensation [10].Fig. 3: T-representation of the DFIGequivalent circuit in dq reference frameThe doubly fed induction generator ismodeled in reference Park [13,14], leadingto the following equations:VVVVFig. 2: Schematic of Unified Power Quality ConditionerɸR iɸR iɸR iɸR iw ɸw ɸww ɸww ɸ(1)In this paper, the control strategy used for theUPQC is the Artificial Neural Networks (ANN); itis a new control approach, used in manyengineering area [11], it has many advantages; firstANN do not need to be reprogrammed, they learnfrom the first program even when an error occurs inthe network; they continue working by dint of theirparallel nature, the ANN can be implemented easilyin any application without any problem, theydepend on the database and prototype [12] withouta mathematical analysis.Therefore, the main contribution of this paper is tostudy the efficiency of ANN approach applied tothe UPQC when the DFIG wind turbine connectedto the electrical grid under sag, swell and harmonicsconditions.In the proposed work, the first section presents amodeling of the DFIG, the UPQC and the systemunderstudy are discussed in the second section, theANN control strategy is explained in the thirdsection, the Matlab/Simulink simulation resultsunder different conditions are discussed in thefourth section, finally the last section is aconclusion.[ɸ , ɸ ], [ɸ , ɸ ]: the componentsof the flux of the stator and the rotor[V , V ], [V , V ]: the components ofthe voltage of the stator and the rotor[i , i , [i , i : the components of thecurrent of the stator and the rotorRs, Rr: the stator and rotor resistancerespectively.The stator and rotor inductance (Ls, Lr) aregiven in equation 2,(L , L ): the leakage inductanceL : the magnetizing inductanceLLLLLL(2)The flux linkages are given in equations 3:5978

Journal of Theoretical and Applied Information Technology15th September 2018. Vol.96. No 17 2005 – ongoing JATIT & LLSISSN: 1992-8645ɸɸɸɸwww.jatit.orgLiLiL iL iL iL iLiLiThe topology of the UPQC is shown in fig. 4, itconsists of 2 voltage source inverters (VSI) whichare connected back to back with each other, theyshare a common DC link, the first VSI acts as ashunt active power filter (APF), the other as a seriesone [15](3)The active and reactive power equations atthe stator and rotor windings are:3V iV iP2QV iV i(4)3.E-ISSN: 1817-3195The series converter imposes a sinusoidal current inphase with power supply through the couplingtransformer, while the shunt converteracts as a current source with high value impedance,it isolates the grid from the currents harmonic.Both shunt and series APF are IGBT based threeleg bridge inverters, at the output of the APFsinterfacing inductors are used, also to filter highfrequency high pass RC filter are used.CONFIGURATION OF THE UPQC :Fig. 4: UPQC and strategy control topology𝑉In order to attenuate harmonics produced by thewind turbine (WT), the shunt inverter injects acurrent as the following equation:𝑖𝑖𝑖𝑉𝑉(6)Where 𝑖 , 𝑖 , 𝑖 are respectively the seriesinjected voltage, the WT voltage and the supplyvoltage(5)Where 𝑖 , 𝑖, 𝑖 are respectively the shuntAPF current, reference supply current and WTcurrentThe series APF of the UPQC injects voltageexpressed by equation:The UPQC eliminates the harmonics in the supplycurrent, therefore the quality current is improved,the voltage and current are in phase, thus the factorcorrection of the system is improved and there is noneed for an additional equipment to compensate thereactive energy.5979

Journal of Theoretical and Applied Information Technology15th September 2018. Vol.96. No 17 2005 – ongoing JATIT & LLSISSN: 1992-86454.www.jatit.orgProposed control strategy:In this paper; the performance of theUPQC is enhanced by using the ANN controlstrategy. The advantage of the ANN control is afast-dynamic response and the maintain of systemstability. The ANN is interconnected artificialneurons, it has three layers, the first one acts asinput neurons; it sends data to the second layerwhich in turn sends orders to the third output layer;so a defined topology characterizes the ANN [16].The instruction of ANN is to control the shunt APF;the reference signal generation to balance the dclink voltage.E-ISSN: 1817-3195The reference voltage (400 V) is compared with(Vdc) the actual voltage to balance thecapacitor voltage. The input data is thecorresponding error, the number of hidden layers is20, the algorithm used for the network training isGradientDescentwithMomentum,theSimulink/matlab model of the ANN for thecapacitor voltage in the shunt controller of theUPQC is shown in fig. 5The NN is trained in order to afford fundamentalreference voltage, the active power is estimated byapplication of 3 phase PLL to Vacb (the DFIGvoltage), the control strategy used is the StationaryReference Frame with its ɖ-β components, theactive power estimation is shown in fig. 6Fig. 5: ANN structure for capacitor voltage balancing5980

Journal of Theoretical and Applied Information Technology15th September 2018. Vol.96. No 17 2005 – ongoing JATIT & LLSISSN: 1992-8645www.jatit.orgE-ISSN: 1817-3195Fig. 6: Active power estimationFrom values of active power estimation and therefence voltage trained by the ANN, it resultscurrents signals that are compared in a hysteresisband current, then the switching signal of the VSIare provided (fig. 7)Fig. 7: ANN trained voltage and active power estimationFor the proposed control in the series APF of theUPQC, a voltage vector strategy is proposed, thegrid voltage Vabc is vector controlled.5.SIMULATION RESULTS:ParametersThe topology of the proposed system isshown in fig. 4, the parameters of the DFIG, shuntand series resistance and inductance, the source aredescribed in table 1ValueDFIG1,4 MWCapacitorC 5000 μFResistanceSeriesRgrid 0.1ΩInductanceSeriesLgrid 3 mHResistanceShuntRL 0.1Ω5981

Journal of Theoretical and Applied Information Technology15th September 2018. Vol.96. No 17 2005 – ongoing JATIT & LLSISSN: 1992-8645www.jatit.orgInductanceShuntLL 3 mHSourceVoltage(phase-phase)100 VrmsE-ISSN: 1817-3195Table. 1: Parameters of the UPQC system connected tothe DFIG1.1. Operating condition type: NormalFig. 7: Compensation using the ANN UPQC; Vsabc source voltage, Viabc the DFIG voltage, Vinj injected voltageFig. 8: Compensation using the ANN UPQC; Isabc source current, Iiabc the DFIG current, Iinj injected currentAt instant t 0.1s, the converters based UPQCbecomes active and compensate the voltage andcurrent harmonics, the DFIG voltage and sourcecurrent are shown in fig.8 and fig.9.Before the instant t 0.15s the DFIG current washigh 3 times of the rated current, then the shuntconverter acts on the signal of the system throughthe ANN control strategy, a convergence ofcurrents is observed due to the rapid charging of5982

Journal of Theoretical and Applied Information Technology15th September 2018. Vol.96. No 17 2005 – ongoing JATIT & LLSISSN: 1992-8645www.jatit.orgcapacitor, from instant t 0.25s the signals currentsbecome more compensated.The voltage compensation starts at instant t 0.1 sand current compensation starts at t 0.15 s.From fig. 8 it is noticed that the DFIG voltage iscompensated, the harmonics in DFIG current isremarkably reduced.E-ISSN: 1817-31951.2. Operating condition type: under voltage SagFig. 9: Compensation using the ANN UPQC during sag condition; Vsabc source voltage, Viabc the DFIG voltage, Vinjinjected voltage5983