Detecção de Anomalias em Painéis Fotovoltaicos por Meio da Termografia e Inteligência Artificial

Abstract

Photovoltaic modules are the most exposed elements of a hotovoltaic system, being subject to various types of anomalies such as: manufacturing defects, problems with storage and handling or environmental factors, requiring care to avoid loss of efficiency beyond that predicted by the manufacturers. The early diagnosis of anomalies in the panels and their corrections are of great importance for the conservation of its electrical energy production, safety and maintenance of its economic viability. Work focusing on automating the process of identifying anomalies in photovoltaic modules has been developed and published. The presente study proposes a neural network model for analyzing thermographic images of panels installed in operating plants, with the purpose of increasing efficiency in diagnosing anomalies and assisting in decision-making by operation and maintenance teams. To this end, anomalies were grouped in a way not observed in related works in classes that require the same type of intervention, also carrying out a comparison of results with pre-trained models. The proposed model presents an accuracy of more than 90% in the classification of anomalies in thermal images of photovoltaic panels.