An efficient power extraction using artificial intelligence based machine learning model for SPV array reconfiguration in solar industries

Abstract

In the field of solar industries, extracting maximum power from solar photovoltaic systems under partial shading conditions has gained significant attention in recent years. One of the most efficient concepts brought out to extract the power output of the existing solar photovoltaic system is the reconfigurable solar photovoltaic system. Numerous static and dynamic reconfiguration techniques are mentioned in the literature. This work proposed a Modified Sudoku reconfiguration based on static techniques and compared to the most common existing configuration i.e., Total-Cross-Tied and Sudoku photovoltaic array configurations. In addition, an Artificial Intelligence-based machine learning model (Fuzzy Expert System) is implemented for the prediction of suitable configurations for solar photovoltaic arrays under partial shading conditions. The performance of the FES model is evaluated by comparing the predicted results and the results obtained from the simulations on 5 × 4, 6 × 4, 6 × 6 and 9 × 9 PV arrays. Results demonstrated that the implemented FES model generated accurate results and 0% MAPE in all 33 sample cases for predicting the best suitable configuration. This realistic, simple, and cost-effective fuzzy model can be utilized to replace existing estimation systems that employ the use of complex technological analysis and simulation models. Thus, this approach can be very effectively used in the solar industry for selecting the configuration for installing solar panels.