Performance Analysis of Maximum Power Point Tracking of PV Systems using Artificial Neural Networks and Support Vector Machines

Abstract

The study aimed to improve the performance of a photovoltaic (PV) system by optimizing Maximum Power Point Tracking (MPPT) algorithms. The researchers used AI-based techniques such as Artificial Neural Networks (ANN) and Support Vector Machines (SVM) to overcome the limitations of conventional MPPT techniques. To generate the MPP under varying weather conditions, the researchers simulated the Perturb and Observe (PO) based MPPT algorithm. The algorithm was simulated for varying irradiation(G) and temperature(T), and the voltage and current corresponding to the MPP points were recorded. This dataset was then used to train the ANN and SVM. Using the MATLAB ANN toolbox, an ANN-based MPPT algorithm was designed, and its performance was analyzed using mean square error (MSE) in predicting the MPP. The MSE obtained for the ANN was 0.5. The SVM was trained using the regression learner app in MATLAB to predict the MPP, and the cubic SVM algorithm proved to be the most effective in predicting the MPP, with an MSE of 0.15926. The results suggest that the SVM MPPT is more effective in tracking the MPP during varying weather conditions. Overall, the study provides evidence that AI-based techniques such as ANN and SVM can be used to optimize MPPT algorithms and improve the performance of PV systems, particularly under varying weather conditions.