Machine Learning-Based Fault Detection for Enhanced Reliability in Photovoltaic Systems

Abstract

This paper presents a novel approach for fault detection in photovoltaic (PV) systems using an Artificial Neural Network (ANN) model, designed to improve operational reliability and enhance power generation efficiency. As PV systems are increasingly integrated into the power grid, early and accurate fault detection becomes critical to maintain consistent energy production and avoid system damage. Traditional fault detection methods, such as threshold-based monitoring technique, often struggles with adaptability and precision due to variations in environmental conditions and system configurations. In this work, we propose an ANN-based fault detection method that addresses these challenges by leveraging the ANN’s ability to learn and adapt to complex, nonlinear relationships between system parameters. The proposed model is trained using historical data from PV systems, including current, voltage, temperature, and irradiance, under both normal and faulty operating conditions. By learning patterns associated with various types of short circuit faults, the ANN model can distinguish between normal variations and true fault occurrences. The results demonstrate that the ANN-based method achieves high accuracy in identifying faults under diverse conditions, outperforming traditional fault detection approaches in terms of response time and fault classification precision. This approach not only enhances the fault detection capabilities of PV systems but also minimizes maintenance costs and downtime by enabling predictive maintenance. The study underscores the potential of ANN models in advancing fault detection methodologies for PV systems, supporting their efficient and reliable operation in sustainable energy infrastructures.