Fault detection and prediction in grid-connected photovoltaic (PV) systems

Abstract

Increased productivity of photovoltaic (PV) renewable assets can be achieved through the development of advanced software algorithms for identifying and predicting inefficiencies in assets. The scope of this work is to present the development of a data-driven tool capable of detecting anomalies and failures in PV systems, while also providing prediction of future underperformance incidents at the inverter level. The developed tool was experimentally validated on a 1.8 MWp PV power plant installed in Greece. The results demonstrated the effectiveness of the tool in providing valuable insights and accurate failure predictions for common incidents. More specifically, the developed algorithms for predicting data issues and performance faults achieved an accuracy of 96.4%, when predicting errors up to 7 days in advance. This software tool can be used for online monitoring of performance and anomalies in PV systems, requiring only the available historical monitored data and the maintenance log of the test PV system as input parameters.