Deep Learning Image Classification Models for Solar Panels Dust Detection

Abstract

Solar panels, the primary components of solar photovoltaic systems, play a pivotal role in converting sunlight into electricity. However, the efficiency and performance of solar panels can be significantly influenced by environmental factors, notably the accumulation of dust and debris on their surfaces. This paper focuses on the investigation of deep learning image classification techniques to detect dust periodically, utilizing solar panel images collected by drones or robots. This approach aims to reduce the impact of dust on solar panels and help identify effective cleaning methods for each case. This work proposes the development of a deep learning binary image classifier model specifically designed to differentiate between “dusty” and “clean” solar panels. The proposed system is based on pre-trained deep learning models fine-tuned for dusty solar panel detection. The results demonstrate that fine-tuning the weights of the pre-trained model enhances performance, with the EfficientNetB7 model yielding the best outcome.