Smart fault isolation and diagnosis in PV setups utilizing random forest classifiers

In the modern era, there has been a growing focus among researchers on the transition from fossil fuels to renewable energy sources, particularly photovoltaic (PV) energy, which is gaining popularity worldwide. As the development and installation of PV systems accelerate globally, it is essential to address the various faults and failures these systems may encounter. Consequently, fault diagnosis and evaluation have emerged as critical areas of study aimed at enhancing performance, improving system efficiency, and reducing maintenance costs and repair times. This paper proposes the use of a Random Forest classifier (RF) for diagnosing short circuit and open circuit faults in PV systems. The classifier is trained using machine learning algorithms to accurately identify different fault types based on real measured data from an experimental PV setup. This data encompasses weather conditions such as cell temperature and solar irradiation, as well as system parameters like current and voltage at the maximum power point, alongside performance metrics. The Random Forest classifier serves as a proactive tool for maintenance and fault diagnosis in PV systems, contributing to better overall performance and reliability. Testing on real-world data from a PV system demonstrates that this approach achieves remarkable accuracy in fault diagnosis, with a precision of 100% for current classification and around 97% for voltage classification, all within a few seconds for each parameter.