Performance Evaluation of Fault Detection Algorithm Applied on a Photovoltaic Generator

The purpose of this article is to test the effectiveness of the proposed diagnostic method, to detect sensor faults in an PV generator (PVG), in order to ensure better maintenance. The diagnostic method built, is based on the parity space method, to generate residuals. These residuals, compiled by dynamic cumulative sum (DCS) tests, generate different detection functions, which are compared to a decision threshold. The analysis of the performance will be done through the curve of the operational characteristics of reception (ROC) to adjust the rate of detection of failures compared to false alarms. This makes it possible, on the one hand, to optimize the adjustment of the width of the sliding windows used during the statistical test and, on the other hand, to make a precise decision on the choice of the detection threshold. The quality of the detection and location of sensor faults inherent to the optimal operation of the PVG, makes it possible to undertake the appropriate maintenance actions.