Intelligent Condition Monitoring in Photovoltaic Systems: A Lightweight Multi-Scale Time Series Neural Network

In the context of a complex and dynamic external environment, photovoltaic systems are susceptible to a range of failures. Currently, conventional monitoring systems and fault detection techniques fall short of meeting the advanced intelligent requirements of these systems. We proposed a lightweight recurrent neural network-based condition monitoring approach with multi-scale time series, which achieved rapid and accurate anomaly detection under mixed mismatches in photovoltaic systems. Firstly, this study examines the diverse mismatch phenomena observed in photovoltaic arrays. Then, by extracting multi-scale time series features of photovoltaic system operation data, an implicit mapping relationship between the operational data and the mismatch conditions is established. Finally, the effectiveness of the method was verified through self-made photovoltaic power stations.