A Comparative Analysis of ANN Algorithms Performance for Maximum Power Point Tracking in Solar Photovoltaic System

Abstract

This paper presents the comparative analysis of Artificial Neural Network (ANN) based algorithms in maximum power point tracking (MPPT) for solar photovoltaic system. The algorithms deployed in this paper are Bayesian Regularization (BR), Levenberg- Marquardt (LM) and Scaled Conjugate Gradient algorithm (SCG).The MPPT model for solar photovoltaic system was designed in MATLAB/Simulink environment and ANN toolbox was used to for analysis. For training 70% data was used and rest 30% was used for validation and testing purpose, which was 15% each. The proposed model was trained seven times for each algorithm and best result was taken. The performance of BR algorithm was better in terms of mean square error which was less than LM algorithm. But with LM algorithm the learning rate, thus time required for training is less so it can be preferred over Time constrained system. SCG algorithm trained the system perfectly with low performance hence it is not suitable for MPPT module. Solar module of 200W with 2 modules in series and 1 module in parallel were taken. The output generated from the trained MPPT solar energy system was 400 W.

Introduction

I. INTRODUCTION

The block diagram consistof PV cell, DC-DC boost converter, PWM generator and with ANN controller is shown in Fig.1 [6]. It consist of  Artificial Neural Network (ANN) having input G (radiance),T(temperature), Vpv (voltage across PV module), Ipv (current from PV module), PWM generator which will generate duty cycle D. It is also an input for DC-DC boost converter. The converter output is connected to load. The voltage across load can be changed by changing the duty cycle. While the duty cycle is decided from  ANN.

For the implementation of ANN three techniques has been used namely Bayesian Regularization, Levenberg- Marquardt and Scaled Conjugate Gradient algorithm. A comparative analysis about performance of these algorithms has also been doneunder different constraints.

The PV module has following components PV array, Artificial Neural Network based MPPT, PI controller, PWM Generator, DC-DC boost Convertor shown in Fig 1.

For inputs of PV array Solar Irradiance is taken in W/m2 and temperature is taken in degree in degree Celsius. The data is collected from the Lucknow City in India (26.84oN , 80.94o E) to get hourly average values of solar energy dataset. Depending upon the data set a plot between hourly temperature (T) and irradiance (G) is shown in Fig. 2.

The artificial neural network based MPPT was trained from the given solar energy data. The boost DC-DC convertor is used in a circuit which is controlled by duty cycle. Duty cycle is generated from the PWM generator from output of difference between PV module voltage and voltage output from artificial neural Network based MPPT algorithm.

II. POWER CIRCUIT

It consists of PV cell and DC-DC boost converter. For detailed analysis of power circuit small signal model of PV cell is already available in literature [15].

A. PV cell

PV cell's  model is shown in Fig. 3 [15]. It consists of one current source, one photo detector diode and two resistances. The cell photocurrent or current source is represented by the Ip. The current through the diode is represented by id. The intrinsic shunt and series resistances of the cell are denoted as Rsh and Rs, respectively. Since Rsh and Rs typically have very large values, respectively, it is possible to ignore them to make the analysis simpler. Practically speaking, PV cells are arranged into bigger units called PV modules, and these modules are linked together in series or parallel to create PV arrays.

.

B. DC-DC Boost Convertor

After PV module the output is fed to DC-Dc boost converter. It is required for enhancing the output voltage obtained from PV cell. The schematic diagram of boost converter is shown in Fig. 4 [8].  It consist of one diode, inductor Ls, input Capacitor Ci , output Capacitor Co and a controlled switch can be MOSFET or IGBT.

III. CONTROLLING CIRCUIT

For controlling of DC-DC boost converter Maximum Power Point Tracking (MPPT) algorithm has been used. MPPT technique is implemented through ANN. Three algorithms Levenberg Marquardt, Bayesian Regularization and Scaled Conjugate Gradient is used.[6] Comparative analysis has been done for these three methods and discussed in SIMULATION section. To extract the maximum power from PV cell Maximum Power Point Tracking (MPPT) algorithm has been implemented by using these three ANN techniques.

A. Maximum Power Point Tracking (MPPT)

To extract maximum power from the solar panel MPPT is used. In a photovoltaic module maximum power can be drawn from a single operating point at any given time [3]. This operating point is denoted as Maximum Power Point Tracking (MPPT).The core concept of maximum power point tracking is to maintain the operating point of photovoltaic module at maximum power point [8].To get maximum Power Point the line peak power point in P-V curve and load line intersection in I-V curve should be on a same line. Since output power of solar panel is dependent on irradiance and temperature which changes frequently so to extract maximum power output at changing irradiance MPPT algorithms are widely used.

DC-DC convertor is used to supply DC supply to load and it is controlled by duty cycle. Various MPPT algorithms are used for tracking maximum power point.           In this research paper we are using neural network based MPPT

B. Artificial Neural Network (ANN)

Artificial Neural Network is a collection of artificial neurons which replicate the human neurons. Input is given to each artificial neuron which produces an output. Each input signal is associated to a weight to give targeted output.[2]

Bayesian Regularization [BR] minimise a linear combination of square errors and weights of the neural network model. Linear combination is also modifieds for better quality of training. BR algorithm does not require cross-validation.[6]

Overtrain the BR algorithm is difficult since evidence procedures enable a Bayesian objective criterion for terminating training. Because the BRANN calculates and trains on a number of effective network characteristics or weights, essentially turning off those that are not significant, they are also challenging to overfit. Typically, this effective value is much lower than the weights in a typical fully connected back-propagation neural network.

Scaled Conjugate Gradient algorithm is based on conjugate directions, but no line search is performed at every iteration. So SCG was designed to reduce computational cost. Thus it reduces memory requirement. SCG algorithm is explained in depth in [16].

IV. SIMULATION RESULTS

A prototype has been developed in MATLAB for implementation of the ANN based algorithm. For training of data using ANN algorithm neural network toolbox is used. Fig.  shows the proposed schematic diagram in MATLAB. Boost converter is used as DC-DC converter. Table I shows the details of components used for the simulation results. Kyocera Solar KC200GT model is used in this research paper. Two modules in series and 1 in parallel configuration is taken. The parameters of solar module are given in Table II.

Table I .Boost Convertor Parameters                                            Table II. Solar Module Data

The input data are irradiance and temperature were fed to the ANN. 1500 samples of irradiance and temperature were collected from location 26.8o N 80.9o E located in India. Output from the trained neural network is Voltage data. Comparative analysis for Bayesian Regularization, Levenberg- Marquardt and Scaled Conjugate Gradient algorithms has been done using NN toolbox. The result from each algorithm is compared to show their application in solar energy. For measuring performance and accuracy of these algorithms parameters such as Mean Square Error (MSE), Gradient, Momentum parameter, regression data, validation check are used.

Gradient is defined a calculation for tuning parameter of ANN model in a manner such that output deviation can be minimised. Epoch is defined as single cycle for training data; it is each trail to train from datasets. Momentum (Mu) is used to avoid local minima problem leading to no convergence. Larger Mu may lead to fast convergence. Validation check is used to reduce model overfitting. Regression is ANN is learning relationship between input and output of the model.

B. Levenberg –Marquardt (LM)

Training of ANN based MPPT using LM algorithm was successful as regression is equal 1 as shown in Fig  .It means that there was perfect prediction of output voltage based on input data and its corelation with output. The best validation performance of LM based ANN is ---- at 377 epoch, means the performance of LM is good for training MPPT based ANN. In Fig.  the zero error in histogram lies at -3.2 e-06. Gradient of LM is 9.9914e-08 which prove the convergence of result is satisfactory.Value of Mu for LM is 1e-08 as shown in Fig. 7.

B. Bayesian Regularization (BR)

The best training performance using BR algorithm is 2.57e-13  with epoch at 469 proves that BR algorithm provides better training than LM with trade of with time taken during training which is higher in BR. Fig.-- . The gradient amd momentum parameter in BR is 9.78e-8 and 500000 respectively shows that learning rate is slower in BR when comparing it with LM.Error histogram shows that the zero error is at -4.6e-08l. Validation error is zero in BR trained network as shown in Fig---. Output from ANN and the target matches perfectly as value of R is 1 as shown in Fig.8

???????

Conclusion

The Performance of Levenberg Marquardt, Bayesian Regularization and SCG algorithms were analysed for MPPT based solar power system. The neural network based MPPT trained with BR algorithm gives better performance with least mean square error but it is at the cost of higher training time whereas LM algorithm provides faster learning rate leading to lower training time.So there is trade-off between performance and time cost for training. If there is time constraint in project then LM algorithm should be preferred over BR. SCG algorithm is not suitable for training of neural network for MPPT. The output power generated at 1000 Irradiance at 250 C temperature is 400 W. MPPT based solar panel generated maximum power with every variation in irradiance value thus extracting maximum power from solar power system under varying weather conditions.

References

REFERENCES [1] Rajib Baran Roy, Md. Rokonuzzaman , Nowshad Amin “A Comparative Performance Analysis of ANN Algorithms for MPPT Energy Harvesting in Solar PV System” IEE Vol. 9, Issue 3 June 2021. [2] Catalina Gonzolez, Castano Carlos, “MPPT Algorithm based on Artificial Bee Colony for PV system” ,IEEE, Vol. 9, Issue 4, Feb 2021. [3] Jerome Cros, Rajib Baran “Maximum Power Tracking by Neural Network” , in Proc. ICRITO IEEE, Vol. 8, p. 257, June 2020. [4] Sabir Messati, Abd Ghani Harg \" New Neural Networks MPPT controller for PV “ Proc. 2nd IREC, Vol 2 , September 2015. [5] Zarko Zecevic , Maja Rolevski “Neural Network Approach to MPPT Control and Irradiance Estimation” Applied Science MDPI, Vol 10, Issue 15, July 2020. [6] Kian Jazayeri, Moein Jazayeri and Sener Uysal, “Comparative Analysis of Levenberg-Marquardt and Bayesian Regularization in photovoltaic power estimation using Artificial Neural Network”, Springer, LNAI 9728, pp 80-95, 2016. [7] R. Algarín, , D. R. Leal and D. S. Hernández “A low-cost maximum power point tracking system based on neural network inverse model controller”, Electronics, vol. 7, p. 244, Jan. 2018. [8] Brigitte Hauke, “Basic Calculation of a Boost Converter’s Power Stage”, Texas Instruments, US, pp. 1-7, Jan 2014. [9] L. Chen and X. Wang, ``Enhanced MPPT method based on ANN-assisted sequential Monte-Carlo and quickest change detection\'\', IET Smart Grid, vol. 2, Issue. 4, September 2019. [10] Henery P. Gavin “ The Levenberg-Marquardt algorithm for Non linear square Curve Fitting Problems” 3rd Edition, Duke University, 2018. [11] Ahmed Gundogdu, Resat Celikel “ANN Based MPPT Algorithm for Photovoltaic Systems, Turkish Journal of Science and Technology, Vol. 15, Issue 2, pp. 101-110, Aug 2020. [12] Samer Gowid, Ahmed Massoud, “A Robust Experimental-based Artificial Neural Network Approach for Photovoltaic Maximum PowerPoint Identification considering Electrical, Thermal and Meteorological Impact. Vol 15, Issue 9, pp 3699-3707, June 2020. [13] M. A. Elgendy, D. J. Atkinson “Experimental investigation of the incremental conductance maximum power point tracking algorithm at high perturbation rates,\'\' IET Renewable. Power Gener., vol. 10, no. 2, pp. 133-139, 2016. [14] Prof. L. Umanand ,“Design of photovoltaic System”. NPTEL , IISC Bangalore, 2018. [15] Jyothi Laxmi P.N “ An Artificial Neural Network Based MPPT Algorithm For PV System”, 4th ICEES, IEEE , PP 237-245 , Feb 2018. [16] Sasmita Nayak, Dr. Neeraj Kumar, “Scaled Conjugate Gradient Back propagation Algorithm for Selection of Industrial Robots”, IJCA, Vol. 7, Issue, 6, Nov 2017.

Copyright

Copyright © 2022 Vashu Dixit, Preeti Agrawal Mittal. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.