A Comparison of Machine Learning-Based Methods for Fault Classification in Photovoltaic Systems

Photovoltaic (PV) energy use has been increasing lately and, being highly dependent on environmental variables, its efficiency becomes a major factor for concern. Additionally, these systems can be affected by several kinds of faults, which can lead to a severe energy loss. In this sense, this work compares different machine-learning-based methods, such as K-Nearest Neighbors (k-NN), Decision Trees (DT), Support Vector Machines (SVM), and Artificial Neural Networks (ANN), for detecting the following faults that can occur in Photovoltaic (PV) systems: Module short circuit, MPPT fault, Open Circuit, Partial Shading, and Degradation. The accuracy and computational time taken for training each classifier were compared. ANN achieved the best accuracy, with 99.65%, while being the slowest to train. The SVM achieved a similar result, with significant less training time. There is lack of discussion on the analysis and comparison of PV fault classification methods in the literature, specially with the focus on further practical applications and computational complexity. This way, those points are the main contributions of this work, along with making all simulations and codes publicly available.