6
Simulation model of ANN based maximum power point tracking controller for solar PV system Anil K. Rai a,n , N.D. Kaushika b , Bhupal Singh a , Niti Agarwal b a Department of Electrical & Electronics Engineering, Ajay Kumar Garg Engineering College, Ghaziabad 201009, India b School of Research & Development, Bharati Vidyapeeth College of Engineering, A-4 Paschim Vihar, New Delhi 110063, India article info Article history: Received 9 January 2010 Received in revised form 21 October 2010 Accepted 21 October 2010 Available online 18 November 2010 Keywords: Controller ANN tracker PV array Cell temperature abstract In this paper the simulation model of an artificial neural network (ANN) based maximum power point tracking controller has been developed. The controller consists of an ANN tracker and the optimal control unit. The ANN tracker estimates the voltages and currents corresponding to a maximum power delivered by solar PV (photovoltaic) array for variable cell temperature and solar radiation. The cell temperature is considered as a function of ambient air temperature, wind speed and solar radiation. The tracker is trained employing a set of 124 patterns using the back propagation algorithm. The mean square error of tracker output and target values is set to be of the order of 10 5 and the successful convergent of learning process takes 1281 epochs. The accuracy of the ANN tracker has been validated by employing different test data sets. The control unit uses the estimates of the ANN tracker to adjust the duty cycle of the chopper to optimum value needed for maximum power transfer to the specified load. & 2010 Elsevier B.V. All rights reserved. 1. Introduction The efficient operation of solar PV system requires an optimum transfer of energy generated in the array to load offered by the combination of battery bank and load. This optimum load require- ment varies with insolation and solar cell’s operating temperature. The load offered by battery to PV array also varies with the state of the charge of the battery and its temperature. The condition for maximum energy transfer from PV array to the battery bank can be achieved by inserting an intermediate dc–dc converter [1–3]. The dc–dc converter (chopper) must continuously adjust the voltage and current level to match both the variable PV output and the load. The schematic diagram of such a system is shown in Fig. 1. The unit comprising a dc–dc converter and a controller is sometimes referred to as the maximum power point tracker (MPPT). Different methods of peak power tracking schemes had been proposed in the past using different control strategies [4–7] some use conventional PID controllers [8] whereas others use rule based fuzzy logic tracking regulators [9]. Review of maximum power point tracking algorithms for stand-alone photovoltaic systems has been presented [10]. These techniques vary in complexity, sensors required, convergence speed, and cost, range of effectiveness and implementation hardware. Recently the artificial neural network (ANN) approach has generated new interest in electrical power control applications. In this paper we have applied a neural network approach, which is well suited for micro controller implementation. In this technique the input variables to converter are currents and voltages of the solar panel corresponding to a given solar radiation and operating cell temperature conditions. The measured solar radiation and estimated module temperature are fed to the ANN tracker; trained and tested offline to yield maximum power voltage (V max ), maximum power current (I max ) and required optimum load corresponding to maximum power transfer. The control unit is used to estimate and adjust the duty cycle of the chopper to optimum value needed for maximum power transfer to the specified load. Essential input data are the values of load resistance, ambient temperature, wind speed and solar radiation. The main purpose of this paper is to develop an ANN based maximum power point tracking controller that extracts maximum power from the solar panel via the chopper under variable radiation and weather conditions. 2. Model of solar PV array The solar PV array is formed by an appropriate series-parallel combination of solar cells that provides the required rated output voltage and current under normal conditions. A solar cell is basically a p–n junction semiconductor that directly converts solar energy into electricity. The solar cell terminal current can be expressed as a function of photo-generated current, diode current and shunt current. The photo-generated current (I ph ) depends on both irradiance and temperature. It is measured at some reference conditions such as reference temperature T c,ref , reference radiation G ref and reference Contents lists available at ScienceDirect journal homepage: www.elsevier.com/locate/solmat Solar Energy Materials & Solar Cells 0927-0248/$ - see front matter & 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.solmat.2010.10.022 n Corresponding author. E-mail addresses: [email protected] (A.K. Rai), [email protected] (N.D. Kaushika). Solar Energy Materials & Solar Cells 95 (2011) 773–778

Simulation model of ANN based maximum power point tracking controller for solar PV system

Embed Size (px)

Citation preview

  • um

    Ag b

    lege,

    schi

    n m

    n de

    ate

    arra

    amb

    tern

    set

    ura

    the

    r ma

    systemarrayd. Thisr cellsay alsomperarray tote dcntinuoariable

    control applications. In this paper we have applied a neural

    rallelutput

    energy into electricity. The solar cell terminal current can be

    Contents lists available at ScienceDirect

    .els

    Solar Energy Mater

    Solar Energy Materials & Solar Cells 95 (2011) 773778as reference temperature Tc,ref, reference radiationGrefand [email protected] (N.D. Kaushika).expressed as a function of photo-generated current, diode currentand shunt current.

    The photo-generated current (Iph) depends on both irradianceand temperature. It is measured at some reference conditions such

    0927-0248/$ - see front matter & 2010 Elsevier B.V. All rights reserved.

    doi:10.1016/j.solmat.2010.10.022

    n Corresponding author.

    E-mail addresses: [email protected] (A.K. Rai),implementation hardware. Recently the articial neural network(ANN) approach has generated new interest in electrical power

    voltage and current under normal conditions. A solar cell isbasically a pn junction semiconductor that directly converts solaruse conventional PID controllers [8] whereas others use rule basedfuzzy logic tracking regulators [9]. Review of maximum powerpoint tracking algorithms for stand-alone photovoltaic systems hasbeen presented [10]. These techniques vary in complexity, sensorsrequired, convergence speed, and cost, range of effectiveness and

    2. Model of solar PV array

    The solar PV array is formed by an appropriate series-pacombination of solar cells that provides the required rated oThe schematic diagram of such a system is shown in Fig. 1. The unitcomprising a dcdc converter and a controller is sometimesreferred to as the maximum power point tracker (MPPT).

    Different methods of peak power tracking schemes had beenproposed in the past using different control strategies [47] some

    radiation. The main purpose of this paper is to develop an ANNbased maximum power point tracking controller that extractsmaximum power from the solar panel via the chopper undervariable radiation and weather conditions.1. Introduction

    The efcient operation of solar PVtransfer of energy generated in thecombination of battery bank and loament varies with insolation and solaThe load offered by battery to PV arrthe charge of the battery and its temaximumenergy transfer from PV aachieved by inserting an intermediadcdc converter (chopper) must coand current level tomatch both the vrequires an optimumto load offered by theoptimum load require-operating temperature.varies with the state ofture. The condition forthe battery bank can bedc converter [13]. Theusly adjust the voltagePVoutput and the load.

    network approach, which is well suited for micro controllerimplementation. In this technique the input variables to converterare currents and voltages of the solar panel corresponding to agiven solar radiation and operating cell temperature conditions.The measured solar radiation and estimated module temperatureare fed to the ANN tracker; trained and tested ofine to yieldmaximum power voltage (Vmax), maximum power current (Imax)and required optimum load corresponding to maximum powertransfer. The control unit is used to estimate and adjust the dutycycle of the chopper to optimumvalue needed formaximumpowertransfer to the specied load. Essential input data are the values ofload resistance, ambient temperature, wind speed and solarSimulation model of ANN based maximfor solar PV system

    Anil K. Rai a,n, N.D. Kaushika b, Bhupal Singh a, Nitia Department of Electrical & Electronics Engineering, Ajay Kumar Garg Engineering Colb School of Research & Development, Bharati Vidyapeeth College of Engineering, A-4 Pa

    a r t i c l e i n f o

    Article history:

    Received 9 January 2010

    Received in revised form

    21 October 2010

    Accepted 21 October 2010Available online 18 November 2010

    Keywords:

    Controller

    ANN tracker

    PV array

    Cell temperature

    a b s t r a c t

    In this paper the simulatio

    tracking controller has bee

    unit. The ANN tracker estim

    by solar PV (photovoltaic)

    considered as a function of

    employing a set of 124 pat

    output and target values is

    takes 1281 epochs. The acc

    sets. The control unit uses

    optimum value needed fo

    journal homepage: wwwarwal

    Ghaziabad 201009, India

    m Vihar, New Delhi 110063, India

    odel of an articial neural network (ANN) based maximum power point

    veloped. The controller consists of an ANN tracker and the optimal control

    s the voltages and currents corresponding to amaximum power delivered

    y for variable cell temperature and solar radiation. The cell temperature is

    ient air temperature,wind speed and solar radiation. The tracker is trained

    s using the back propagation algorithm. The mean square error of tracker

    to be of the order of 105 and the successful convergent of learning processcy of the ANN tracker has been validated by employing different test data

    estimates of the ANN tracker to adjust the duty cycle of the chopper to

    ximum power transfer to the specied load.

    & 2010 Elsevier B.V. All rights reserved.power point tracking controller

    evier.com/locate/solmat

    ials & Solar Cells

  • Nomenclature

    A AmpereD duty cycle of the converterDopt optimum duty cycleE electron charge 1.6021019 CEg energy band gap (eV)G actual solar radiation (W/m2)Gref reference radiation (W/m

    2)ID diode current (A)Iin converter input current (A)

    Iscc manufactured supplied temperature coefcient of theshort circuit current (A/K)

    ISH shunt current (A)K Boltzmann constant, 1.381023 J/KRL load resistance (O)RLin input resistance of the converter (O)Rs series resistance (O)TA ambient air temperature (1C)Tc actual operating temperature of cell (K)Tc,ref reference temperature (1C)V voltage

    Vout converter output voltage (volt)

    A.K. Rai et al. / Solar Energy Materials & Solar Cells 95 (2011) 773778774photocurrent Iph,ref and related as follows [11]:

    Iph G

    GrefIph,ref IsccTcTc,ref 1

    whereG is the actual solar radiation (W/m2), Tc the actual operatingtemperature of cell (K), and Iscc the manufactured supplied tem-perature coefcient of the short circuit current (A/K).

    The diode current is given by the Shockley equation

    ID I0 expeVcZKTc

    1

    2

    where Vc is the voltage across diode (V), I0 the reverse saturationcurrent (A), Z the diode ideality factor, Rs the series resistance (O), ethe electron charge 1.6021019 C, andK the Boltzmann constant,1.381023 J/K.

    The reverse saturation (I0) current is given by [11]

    I0 I0,refTc

    Tc,ref

    3exp

    eEgZK

    1

    Tc,ref 1Tc

    3

    whereas the shunt current ISH is given by

    ISH VcRp

    4

    Imax maximum power current (A)In current estimated by ANN tracker corresponding to

    peak power point (A)I0 reverse saturation current (A)Iout converter output current (A)Iph photo-generated current (A)Iph,ref reference photocurrent (A)where Rp is the shunt resistance (O).Eq. (1) can be written as

    I IphI0 expeV IRsZKTc

    1

    V IRs

    Rp5

    The actual cell temperature (Tc) has been found to be dependent onambient air temperature, radiation and wind speed, which has

    Solar PV

    array

    ANN Tracker

    Input data Irradiance Atmospheric temp Wind speed

    Fig. 1. Schematic diagram of Abeen modeled by the following relationship [11]:

    Tc 1C 0:943TA0:028 irradiance1:528 wind speed4:36

    where TA is the ambient air temperature given in 1C, irradiance inW/m2 and wind speed in m/s.

    Employing Eqs. (1)(6), a non-linear equation involving photo-generated current (Iph), diode reverse saturation current (I0), shuntresistance (Rp) and load resistance (RL) is obtained, which has beensolved for Vc to supply an initial assumed value of Vc0.

    The above solar PV array model has been implemented inMATLAB and the models accuracy is also analyzed through acomparison between experimental data and corresponding simu-lated data as shown in Fig. 2. In this gure the solar irradiance leveland ambient air temperature are 1000W/m2 and 25 1C, respec-tively. It is found that the relative errors between experimental andcorresponding simulated values on peak power, peak-power vol-tage and peak-power current are 3%, 4.98% and 1.5%, respectively,which is of the right order of magnitude and is in accordance withthat reported by earlier author [12]. The incident solar radiation haslarger effect on short circuit current, while the effect on open circuitvoltage is ratherweak. The solar PV arraywill generatemore powerwhen the irradiance is higher. With an increasing temperature the

    Z diode ideality factorVmax maximum power voltage (volt)Vc voltage across diode (volt)Vin converter input voltage (volt)Vn voltage estimated by ANN tracker corresponding to

    peak power point (volt)array voltage drops at high voltages. Operating the cell in this regionleads to a power reduction at high temperature.

    3. Model and training of ANN tracker

    An ANN model of maximum power point as shown in Fig. 3 isimplemented with SIMULINK. It has a three-layer feed forward

    Optimum duty cycle

    Chopper Load

    OptimalControl unit

    NN controlled SPV system.

  • perceptron neural network architecture, which has an input layerof two neurons, a hidden layer of nine neurons and an output layerof two neurons. The network receives the outside input, scales it byweights and biases and passes it to the neurons in the next layer.Each neuron in every layer receives its input from the previouslayer. The input and hidden layer neurons are approximated by tansigmoid activation function whereas output layer neurons areapproximated by pure linear activation function. The model usesthe cell operating temperature and radiation as input values andestimates the corresponding maximum power voltage and currentvalues. Using maximum power point current and voltage dataobtained by the solar PV array model at different temperature andradiation the network has been trained and validated using backpropagation algorithm in MATLAB.

    The following parameters are set while training the feedforward neural network:

    training pattern124,learning rate0.001,set error goal0.00001,number of training iterations1281,momentum0.95.

    It was observed that the mean square error is (performancegoal9.99106) is reached well below the set error in 1281epochs. The accuracy of the trainedANNmodelwas validated usingother sets of data, which are different from those used for thetraining process and the mean square error in this case has alsobeen found (5.4968106) to be less than the set error goal. InFig. 4 the actual data values and the ANN-estimated values ofmaximum power current (Imax) and voltage (Vmax) have been

    compared. The values Imax and Vmax correspond to solar intensityvariationbetween100 and1000 W/m2 and the cell temperatures of10, 25 and 40 1C, whose ranges are relevant to practical situation.One can notice the good agreement between all sets of results.

    4. Model of chopper and controller

    The input and output quantities such as voltage and current ofbuck boost converter under the assumption of a loss less circuit arerelated by the following equations [13]:

    Vout DVin1D 7

    and

    Iout 1DIinD

    8

    where Vout is the converter output voltage, volts; Vin the converterinput voltage, volts; Iout the converter output current, A; Iin theconverter input current, A; D the duty cycle of the converter.

    The converter is in buckmode for Do0.5 and in boost mode forD40.5.

    From Eqs. (7) and (8)

    VoutIout

    D2

    1D2VinIin

    9

    A.K. Rai et al. / Solar Energy Materials & Solar Cells 95 (2011) 773778 775Fig. 2. Simulated and experimental IV characteristics of solar cell array at solarirradiation level and ambient air temperature 1000 W/m2 and 25 1C, respectively.Fig. 3. ANN model of the maximum power poFig. 4. Comparison of actual max power current and voltage values with themaximum power current and voltage values obtained by ANN tracker.int tracker implemented with SIMULINK.

  • This equation can be can be written as

    RLin 1DD

    2RL 10

    where RLin is the input resistance of the converter (O) and RL theload resistance (O).

    The controller will adjust the duty cycle of the chopper for theoptimum transfer of the power to the load. This could only bepossiblewhen the input resistance of the chopper becomes equal tothe resistance of the module (source resistance). The resistanceof the module corresponding to the peak operating power pointcan also be obtained by employing the corresponding currentand voltage estimated by the ANN tracker for given conditions(radiation and temperature). The optimum duty cycle (Dopt) of theconverter thus can be related to the voltage and the currentestimated by ANN tracker and the load resistance as follows:

    Dopt 11

    Vn=RLIn

    p 11where Vn and In are the voltage and current estimated by the ANNtracker corresponding to the peak power operating point.

    5. Simulationofmaximumpowerpoint tracker (controller andchopper) using SIMULINK

    A SIMULINKmodel of a complete SPV system as shown in Fig. 5embodying the ANN based maximum power point controller isanalyzed. The model is based on 94 (9 solar cells in series and 4such strings in parallel) solar PV module, feeding power to load

    (ramp load) through the controller (chopper and ANN maximumpower point tracker). This model contains six subsystems, namelySPV_Model, Module_Temp, Normalising_sys, Denormaliser_sys,ANN Tracker_sys and Chopper_sys.

    The SPV_Model subsystem is characterized by Eqs. (1)(5);the Module_Temp subsystem is characterized by Eq. (6). Theinput parameters for the Module_Temp subsystem are incidentsolar radiation; atmospheric temperature and wind speed andoutput parameter is module-operating temperature. The module

    Fig. 6. Maximum power output (Pmax) versus solar radiation intensity.

    A.K. Rai et al. / Solar Energy Materials & Solar Cells 95 (2011) 773778776Fig. 5. SIMULINK model of the complete SPV system.

  • play

    A.K. Rai et al. / Solar Energy Materials & Solar Cells 95 (2011) 773778 777operating temperature and incident solar radiations are fed to the

    Fig. 7. Output voltage, current and power disSPV_Model subsystem. Normalising_sys and Denormaliser_syssubsystems are used to normalize and demoralized the data. TheANN Tracker_sys subsystem gives the maximum voltage andcurrent for all practical values of radiation and temperature. TheChopper_sys subsystem is characterized by Eqs. (7)(11). Thevariation of maximum power output (Pmax) versus radiation isshown in Fig. 6. A ramp load has been considered to show that theoptimum power delivered is constant at particular radiation andtemperature at variable conditions of load. Output voltage, currentand power display of the SIMULINK model of the complete SPVsystem is illustrated in Fig. 7.

    6. Results and discussion

    In this paper an ANN based maximum power point trackingcontroller for Solar PV system has been proposed and analyzed inMATLAB. The work involves the modeling of solar PV array,modeling of chopper and controller and off-line training andtesting of ANN tracker. The solar PV array models accuracy isanalyzed through comparison between experimental data and thecorresponding simulated data. It is found that the relative errorsbetween experimental and corresponding simulated values onpeak power, peak-power voltage and peak-power current are 3%,4.98% and 1.5%, respectively, which is of the right order ofmagnitude and is in accordance with that reported by earlierauthors. The model is used to generate several sets of maximumpower voltage and current data under variable atmospheric andload conditions. The data is used for off-line training and testing ofthe ANN tracker for maximum power voltage and current valuesunder arbitrary input conditions. The optimal control unit com-pares the output of the ANN tracker and the load and adjusts theduty cycle of the chopper such that the chopper operates formaximumpower transfer to the specied load. In this way through

    of SIMULINK model of complete SPV system.the ANNbasedMPPT controller extractsmaximumpower from thesolar array and transfers it to the load. The performance of MPPTcontroller has been incorporated in the complete solar PV systemand analyzed through a SIMULINK model.

    7. Conclusion

    A rigorousmodel for the energy generation by the solar PV arrayhas beendeveloped. Themodel takes into account the effect of solarirradiance, atmospheric temperature, wind speed and variability ofload. The model is used to predict maximum power voltage andmaximum power current under variable atmospheric and loadconditions. Several sets of data for the variability in real physicalconditions as encountered in practice are generated. The resultantdata is used for training the target values of the ANN tracker usingback propagation algorithm. Themean square error of the tracker isset to be of the order of 105 and the successful convergence oflearning process takes 1281 epochs. The peak power trackingcapabilities of theproposed schemebased onANNbasedmaximumpowerpoint tracking controller is demonstrated through simulatedand experimental results. The simulated results show that the ANNbased controller in itsmaximumpowerpoint trackingperformanceexcels over the conventional PID controller and avoids the tuning ofcontroller parameters.

    References

    [1] S.M. Alghuwainem, Steady state performance of dc motors supplied fromphotovoltaic generators with step-up converter, IEEE Transactions on EnergyConversion 7 (1992) 267272.

    [2] J.J. Appelbaum, Starting and steady state characteristics of dc motors poweredby solar cell generators, IEEE Transactions on Energy Conversion 1 (1986)1723.

  • [3] W.Z. Fam, M.K. Balachander, Dynamic performance of a dc shunt motorconnected to photovoltaic array, IEEE Transactions on Energy Conversion3 (1988) 613617.

    [4] Z. Salameh, D. Taylor, Step-up maximum power point tracker for photovoltaicarrays, Solar Energy 44 (1990) 5761.

    [5] M. Veerachary, T. Senjyu, K. Uezeto, Voltage based maximum power pointtracking of PV systems, IEEE Transactions on Aerospace and Electronic System38 (2002) 262270.

    [6] N. Dasgupta, Ashish Pandey, A.K. Mukherjee, Voltage-sensing based photo-voltaic MPPT with improved tracking and drift avoidance capabilities, SolarEnergy Material and Solar Cells 92 (12) (2008) 15521558.

    [7] Kenji Kobayashi, Ichiro Takano, Yoshio Swada, A study of a two stagemaximumpower point tracking control of a photovoltaic system under partially shadedinsolation condition, Solar Energy Material and Solar Cells 90 (1819) (2006)29752988.

    [8] O. Wasynczuk, Modeling and dynamic performance of a line-commutatedphotovoltaic inverter system, IEEE Transactions on Energy Conversion 4 (3)(1989) 337343.

    [9] M. Veerachary, T. Senjyu, K. Uezeto, Feed forward maximum power pointtracking of PV systems using fuzzy controller, IEEE Transactions on Aerospaceand Electronic System 38 (3) (2002) 969981.

    [10] V. Salas, E. Olias, A. Barrado, A. Lazaro, Review of the maximum power pointtracking algorithms for stand-alone photovoltaic systems, Solar EnergyMaterial and Solar Cells 90 (11) (2006) 15551578.

    [11] R. Chenni, M. Makhlour, T. Kerbache, A. Bouzied, A detailed modeling methodfor photovoltaic cell, Energy 32 (9) (2007) 17241730.

    [12] F.C. Treble, Generating Electricity from the Sun, Pergamon Press, New York,1991.

    [13] N. Mohan, Tore M. Undeland, William P. Robbins, Power ElectronicsConverters: Application and Design, John Wiley & Sons (Asia) Pvt. Ltd, 2004.

    A.K. Rai et al. / Solar Energy Materials & Solar Cells 95 (2011) 773778778

    Simulation model of ANN based maximum power point tracking controller for solar PV systemIntroductionModel of solar PV arrayModel and training of ANN trackerModel of chopper and controllerSimulation of maximum power point tracker (controller and chopper) using SIMULINKResults and discussionConclusionReferences