New Method of Detection, Identification, and Elimination of Photovoltaic System Faults in Real Time Based on the Adaptive Neuro-Fuzzy System

Abstract

This article proposes an adaptive neuro-fuzzy system (ANFIS)-based method to detect, identify, and eliminate defects/faults in a photovoltaic (PV) system. The control is conducted using two tests. The first test measures the difference between the power estimated by the ANFIS and the real power of the generator. This test detects defects and isolates the affected branch using controlled switching devices to prevent shutting down the entire power supply by disconnecting the inverter. The second test measures the consistency between the real current and voltage of the generator and the short-circuit current and open-circuit voltage estimated by the ANFIS, to identify open-circuit and short-circuit conditions. Each system has its own dedicated irradiation and ambient temperature sensors whose positions differ to avoid the influence of faults overlap. The method was studied through simulation using MATLAB Simulink and experimentally validated via panels of different power ratings and technological makeup that are more than five years old, utilizing a DS1104 dSPACE controller board. The technique overcomes the limitations of existing methods that are based on comparisons between the estimated and measured variables as the discrepancy between real and simulated behavior; system change because of environment, topology, or age; and the need for frequent intelligent system updates as it does not rely on real data for learning. Additionally, this method includes an aging analysis, and it can be used to control most typical PV installations via a gain factor—the ratio of the new value to the learn value.