A Highly Adaptable Performance Evaluation Method for PV Systems Using Remote Meteorological Data and On-Site Reference Comparisons

This study presents a highly adaptable, sensor-free framework for the performance evaluation of photovoltaic (PV) systems, aiming to address the limitations of traditional methods that rely on expensive on-site sensors. The proposed method includes a two-phase approach. In the training phase, data from 2018 to 2022 were used to derive a combined loss coefficient (Ko) that links global horizontal irradiance (GHIith power output. The evaluation phase uses this coefficient to infer local GHI and calculate the theoretical power output, which is then compared with actual power generation to assess system performance. The framework also employs a peer-comparison strategy as a contingency plan when remote meteorological data is unreliable. By using K-means clustering and the SHAP tool, the method can effectively diagnose the causes of performance degradation, such as shading. The results prove that the proposed framework is a workable and versatile approach for analyzing PV system performance. The peer-comparison method further provides a valuable reference for maintenance teams in data-scarce scenarios.