Leveraging High-Fidelity Sensor Data for Inverter Diagnostics: A Data-Driven Model Using High-Temperature Accelerated Life Testing Data

Inverters pose substantial reliability risks and significantly impact operations & maintenance costs in photovoltaic (PV) systems. Understanding and predicting inverter failure processes is a key enabler for improving levelized cost of energy and competitiveness of the PV industry. In recent years, there has been a growing interest in harnessing sensor information from inverters to monitor and predict inverter degradation and failure risks. In this paper, we propose a comprehensive diagnostics framework for PV inverters that (i) transforms functional sensor information to time-frequency domain features in an effort to capture both summary statistics and signal dynamics, and (ii) uses the produced signal features to build a diagnostic model that predicts degradation severity in PV inverters. Results using inverter data from an accelerated life testing experiment show that proposed approach offers 91-97% accuracy in predicting degradation severity.