Online fault diagnosis for photovoltaic modules based on probabilistic neural network

Photovoltaic (PV) modules often suffer from various faults due to the harsh working environment. This paper proposes a PV fault diagnosis model based on probabilistic neural network (PNN), aiming to enhance the efficiency and reduce the maintenance cost of PV power stations. The influencing factors of PV faults were analyzed in details, and the output features of PV modules under fault states were simulated on MATLAB. Based on the simulation results, the fault types of PV modules were summed up, and a PNN-based PV fault diagnosis model was established. The effectiveness of our model was verified through simulation and experiment and compared with that of the diagnosis model based on backpropagation neural network (BPNN). The results show that our model can effectively detect four types of fault for PV modules, namely, short circuit, open circuit, abnormal degradation and partial shading, and enjoys high accuracy and robustness. © 2019 Lavoisier. All rights reserved.