Detection and Classification of Faults in Photovoltaic System Using Random Forest Algorithm

Abstract

Detection of faults in a photovoltaic system is a great challenge, for increasing the solar power generation and improving efficiency. Under low irradiance condition, the power generation gets reduced, at that time the line-line fault remains undetected. The array current, voltage and irradiance are measured and used for detecting and classifying the fault. This paper proposes a new fault classification algorithm based on supervised machine learning technique. The features are extracted for different test conditions under normal and partial shading conditions to get a sample dataset. The features in the database are analysed using Random forest classification algorithms. The classification accuracy of faults is evaluated using the confusion matrix. The experimental and simulated database results are collected from 100 W PV module with a 4 × 4 array configuration with a successful classification of faults with an accuracy of 99.98%.