Machine Learning Techniques for the Assesment of Open Circuit Voltage Losses in Photovoltaic Systems

This investigation introduces a pioneering approach for remotely monitoring photovoltaic (PV) module open-circuit voltage (Voc) using SCADA data, without incurring energy production losses. By synergistically combining machine learning algorithms for data modeling with fundamental physics-based knowledge, our methodology successfully identifies Voc values from voltage SCADA data and provides a suite of Voc-related performance metrics. These metrics enable the Failure Detection and Diagnosis (FDD) through monitoring deviations in the performance of the module. This FDD strategy allowing for the optimization of the PV operation and intelligent maintenance (O&M) of the asset. The proposed approach is validated using synthetic data, and its robustness is assessed under diverse operational conditions. The findings reported in this study have significant implications in the enhancement of the reliability, efficiency, and overall performance of PV installations, ultimately contributing to the widespread adoption of solar energy as a clean, sustainable, and robust power source.