A Novel Approach to Fault Management Using ANN in a Grid-Tied PV Plant

Installation of a photovoltaic (PV) system in the existing grid is highly followed to meet the increasing electricity demand in recent times. However, this combined grid-tied PV system is susceptible to various types of faults like any other conventional grid. That’s why an effective fault monitoring system plays a crucial role in counteracting the potential damages upon the occurrence of any disturbance. Different types of machine learning (ML)-based approaches have been proposed lately for developing a fault monitoring system considering the necessity. These approaches are designed based on the classification algorithm, which is a supervised learning technique. Most of these works have focused on faults appearing either on the feeders or inside the PV system. Considering this limitation, this study proposes a novel ANN-based approach to detect, classify, and localize faults occurring both inside and outside the PV system. The endeavor begins with designing a detailed Simulink model of a grid-connected PV system. A novel methodology has been proposed to effectively integrate an intelligent system into this model. This system is capable of providing timely responses to any fault, using continuously fed data of irradiance, temperature, and bus currents. The performance of this system proposes a promising solution to redundancy by incorporating a single unit to monitor faults of different parts of a system. The system has been validated for faults occurring at different distances on the grid to identify the range for accurate classification and localization.