FEATURE SELECTION USING SPARSITY-AWARE DIFFERENTIAL EVOLUTION FOR ACCURATE DEFECT DETECTION IN PHOTOVOLTAIC MODULES

Abstract

The efficacy of the photovoltaic (PV) modules is affected by the presence of defects in the modules, therefore monitoring of the PV modules is imperative to ensure reliability and longevity of modules. In this study the features are extracted using histogram of gradients (HOG) and then a novel Sparsity Aware-Differential Evolution algorithm (SA-DE) based feature selection method is introduced for defect classification in electroluminescence (EL) images of PV modules. Several other optimization techniques, such as Genetic Algorithm (GA), Bayesian Optimization (BO), and Ant Colony Optimization (ACO) showed unsatisfactory performance for feature selection. The proposed SA-DE approach exhibits better efficiency in terms of feature selection and defect detection. It is obvious from the results that the method proposed demonstrates superior performance than GA, BO and ACO by achieving an average accuracy of 96 %. A comparative analysis of different optimization methods using Support Vector Machine (SVM) with k-fold cross-validation is also provided to highlight the superiority of the proposed method. Also, the work proposed here is aligned with the United Nations’ Sustainable Development Goals (SDGs), particularly SDG-7 by promoting sustainable energy solutions through improved fault detection in solar modules.