Identification and Classification of PV Array Faults Using Artificial Neural Network

Photovoltaic fault detection and classification are important for an efficient and reliable power supply. The nature of electrical parameters helps in the detection and classification of faults in PV arrays. This research presents an artificial neural network (ANN)-dependent fault detection methodology for detecting line-ground (LGF), hot spot (HSF), and short-circuit (SCF) faults in PV arrays. The proposed technique uses the Neural Network Pattern Recognition methodology to classify the SCF, HSF, LGF, and standard state of operation. A 3.2 kW PV array model is developed on the MATLAB-Simulink platform to verify the proposed work. To validate the suggested methodology, several fault conditions are simulated by varying irradiances, temperatures, and fault resistances. A large enough number of training patterns are produced through simulation, which are then used to train the ANN model. Additionally, performance errors during training help determine the optimum quantity of hidden neurons. Further, the confusion matrix analysis is employed to evaluate the accuracy of the proposed method’s performance. The proposed approach offers a promising accuracy of 99.7% in overall cases.