Using machine learning in photovoltaics to create smarter and cleaner energy generation systems: A comprehensive review

Abstract

Photovoltaic (PV) technologies are expected to play an increasingly important role in future energy production. In parallel, machine learning has gained prominence because of a combination of factors such as advances in computational hardware, data collection and storage, and data-driven algorithms. Against this backdrop, we provide a comprehensive review of machine learning techniques applied to PV systems. First, conventional methods for modeling PV systems are introduced from both electrical and thermal perspectives. Then, the application of machine learning to the analysis of PV systems is discussed. We focus on reviewing the use of machine learning algorithms to predict performance and detect faults, and on discussing how machine learning can help humanity to achieve a cleaner environment in the worldwide drive towards carbon neutrality. This review also discusses the challenges to and future directions of using machine learning to analyze PV systems. A key conclusion is that the use of machine learning to analyze PV systems is still in its infancy, with many small-scale PV technologies, such as building integrated photovoltaic thermal systems (BIPV/T), not yet benefiting fully in terms of system efficiency and economic viability. The wider application of machine learning to PV systems could therefore forge a shorter path towards sustainable energy production.