A Novel Enhancement Integrated Convolutional Neural Network for Automated Defect Detection in Photovoltaic Modules

Abstract

Due to ever-increasing demand of power globally, the focus is now shifted on renewable energy sources like hydro, wind, solar energy from non-renewable energy sources. Solar energy stands is a leading contributor in the sustainable energy revolution. The solar cells which are used to harness the solar energy suffer from several defects arising during the manufacturing or installation processes. Identifying these defects is crucial to prevent degradation in the performance of solar cells. However, manual detection is time consuming and tedious. Hence, automatic defect detection methods must be developed to improve efficiency. This paper proposes a novel image enhancement integrated Convolution Neural Network (IE-CNN) for defect detection in solar cells that involves pre-processing the input image through image enhancement and defect detection using CNN. The hyperparameters of the CNN are selected after extensive experimentation. The method converges fast and also takes care of the overfitting phenomenon. A dataset of electroluminescence (EL) images is used for the implementation of the proposed method. The proposed method achieved a precision of 0.97 and recall of 0.98, culminating in an elevated F1-score of 0.90.