Computer vision and artificial intelligence for anomaly detection in photovoltaic panels

Power generation through photovoltaic technology has become an essential source of electricity in the world in recent years. In Brazil, it is already the second most immense installed power in the electrical matrix, and, as a result, anomalies in photovoltaic panels tend to become a source of losses because they are elements exposed to the actions of nature. Techniques for detecting problems in panels have been studied, and among these, the use of thermographic images collected with the help of drones stands out. In this work, a solution is developed using computer vision and artificial intelligence techniques for detecting and classifying anomalies. It aims to assist operation and maintenance teams in deciding what type of intervention to carry out, thus optimizing their work and mitigating more significant losses. The average result of the Recall metric was over 95% accuracy, demonstrating the viability of its use for developing a solution for the photovoltaic system maintenance market.