Fault Diagnosis Method of Photovoltaic Array Based on Random Forest Algorithm

Abstract

This paper proposes a fault diagnosis method for photovoltaic array based on random forest algorithm. The random forest algorithm used in this paper is based on classification regression tree (CART). And bootstrap sampling method was used to generate multiple training sets from the original training set, and random attributes were introduced in the training process to generate multiple CART trees. When troubleshooting, the final troubleshooting results are determined by voting according to the diagnosis results of each CART tree. By building a photovoltaic array model in Matlab/Simulink simulation module, use it to summarize the output characteristics of photovoltaic array in the common fault states (partial shading, abnormal component aging, open circuit, short circuit). And considering the main natural factors that affect the output of photovoltaic array, the open circuit voltage, short circuit current, maximum power point voltage, maximum power point current, environmental temperature and light irradiance were selected as the input parameters of the random forest algorithm. Then, a 2×2 small photovoltaic array experiment platform and a field data acquisition system were designed and built to collect data in order to further verify the feasibility of the random forest algorithm. The test results of random forest were compared with those of single CART decision tree, K nearest neighbor (KNN) and support vector machine (SVM) algorithms. The results showed that random forest could complete the fault diagnosis task of photovoltaic array better.