From ANN Prediction to Closed-Loop PV Control: A Mini-Review of Thermal Management and Sun Tracking

Abstract

Artificial neural networks (ANNs) are increasingly being used to improve photovoltaic (PV) performance under changing environmental conditions. However, good prediction accuracy does not always lead to effective real-time control. This mini-review examines the use of ANNs in PV thermal management and sun-tracking, with emphasis on how predictions are translated into operating decisions and physical control actions. Recent studies are compared based on their data sources, validation methods, control outputs, energy benefits, robustness, and level of practical implementation. The reviewed approaches are grouped into three stages: prediction, decision support, and closed-loop control. The analysis shows a clear gap between accurate ANN predictions and their reliable use in real operating systems. Many studies remain focused on prediction or optimization, while fewer demonstrate independent validation, real-time hardware operation, cross-site performance, or the energy consumed by cooling and tracking devices. Future work should focus on reliable models, low-cost implementation, net energy benefits, real-time validation, and combined thermal and tracking control. Overall, the key challenge is moving from accurate prediction to reliable and energy-efficient control of PV systems.