A study on fault diagnosis in solar photovoltaic array using artificial neural network

Timely detection, diagnosis and mitigation of short-circuit and arcing faults are essential to avoid catastrophic damage and service discontinuity of solar photovoltaic (PV) arrays. In this work, an application of artificial neural network (ANN) is demonstrated for the fast and accurate diagnosis of string fault, string-to-ground fault and string-to-string fault in a 250 kW grid-tied solar PV array. The electrical features are chosen based on information gain ratio and the model hyperparameters are set based on random search optimization. It is observed that the ANN model displays high classification accuracies for the training (100%), validation (100%) and testing (98%) data; small generalization error (0.02) and fast training speed (7s). Most of these indices are improved compared to the earlier works. Such a smart fault-detection scheme can be deployed physically in a solar power plant.