Detection of Photovoltaic Generator Underperformance Due to Inverter Operation Modes

The increasing deployment of photovoltaic (PV) systems demands advanced supervision tools to ensure optimal performance and minimize maintenance costs. This work presents a novel methodology for detecting and classifying inverter-related anomalies—specifically inverter clipping and thermal power derating—using synthetic SCADA data from a simulated PV plant. The proposed approach integrates linear regression, residual filtering, and Gaussian Mixture Models to identify non-MPP operating conditions, enabling improved accuracy in performance analysis. Results demonstrate high detection accuracy. The lightweight, unsupervised nature of the method makes it suitable for real-time SCADA integration and future application to diverse PV configurations, paving the way for more reliable and condition-based maintenance strategies in PV asset management.