A Degradation‐Responsive Framework for Long‐Term PV Power Estimation

Abstract This paper presents a framework for long‐term photovoltaic power estimation, applicable to all PV systems with at least one year of irradiance and temperature data. After data validation, a static model is trained to replicate initial system performance without degradation. By comparing this model's estimates with real measurements, the baseline efficiency trend is derived, serving as a weighting factor for irradiance to account for performance loss. This dynamic adjustment ensures sustained accuracy over time despite degradation. Tested on five diverse PV systems varying in technology, configuration (single/dual‐tilt), and module type (monofacial/bifacial), the framework maintained high accuracy in long‐term datasets, outperforming the static model as degradation progressed. It also excelled in complex installations, such as dual‐tilt bifacial systems and clipping scenarios, achieving normalized Mean Absolute Errors of 1.1%–2%.