INFRARED THERMOGRAPHY AND AI TECHNIQUES FOR SOLAR PV HOTSPOT DETECTION: A COMPREHENSIVE ANALYSIS OF ML AND DL FRAMEWORKS

Abstract

The presence of hotspots in solar photovoltaic (PV) panels leads to decreased efficiency of the energy produced, quicker degradation of materials and presents
possible safety risks. Hotspots can occur due to shading, defects in the manufacturing process or other fault in the operation of the PV systems. This study proposes
a framework based on machine learning for classification of hotspot areas appear in PV modules using thermal imaging and healthy zones in an experimental investigation. The
process for identifying those areas is based on automated image patching of thermal images and their consequent classification through various machine learning algorithms: Support
Vector Machine (SVM), k-Nearest Neighbours (KNN), Random Forest and Decision Tree models. The success of the models is assessed using explicit measures of effectiveness like
accuracy, precision, recall and F1 score. The results obtained from the classification experiments show that the SVM model gave the highest classification efficiency of 94.03% with
KNN (91.65%) next, then Random Forest (91.17%) and Decision Tree (83.29%). The results demonstrate that the SVM model gave a higher degree of generalization and robustness than
the other classification models for hotspot identification. The proposed algorithm will give a low weight, low computation and non-invasive method of PV fault diagnosis giving improved
maintenance procedures and operational reliability of solar energy generation systems. This article also provides an insight into the latest innovations in utilizing machine learning (ML)
and deep learning (DL) techniques to identify and locate hotspots found within thermal infrared images.