Machine learning applications in health monitoring of renewable energy systems

Rapidly evolving renewable energy generation technologies and the ever-increasing scale of renewable energy installations are driving the need for more accurate, faster, and smarter health monitoring methods. Machine learning (ML) has been widely used for defect identification and fault diagnosis (DIFD) in renewable energy systems (RES) due to its excellent data analysis and pattern recognition capabilities. However, there is still a lack of comprehensive and in-depth research to summarize the progress of ML in RES DIFD. Considering their diversity and evolution, this research systematically reviews the application of ML to DIFD in photovoltaic and wind power systems, providing a detailed analysis of key points, including defects and faults in RESs, available data for DIFD, and ML-based solutions. For each ML application, the model and structure, input and output settings, sample size, equipment technology and scale, target defects/faults, performance, and limitations are discussed. In addition, based on the statistics, analysis, and summary of representative works on RES DIFD, several commonly used ML algorithms are compared. The main trends, challenges, and prospects of applying ML to RES DIFD are extracted and presented. This research is of great significance in guiding further improvements of ML techniques to address the problems of RES DIFD.