Dataset of "AI-based performance analysis of roof-integrated thin-film CdTe photovoltaic system deployed under long-term outdoor conditions"

This paper presents a comprehensive assessment of the long-term performance degradation of a Building-Integrated Photovoltaic (BIPV) system utilizing cadmium telluride (CdTe) modules over a six-year monitoring period. A hybrid methodology combining an Artificial Intelligence (AI)-based predictive model, effective peak power trend analysis, and measured I-V curve evaluations was employed to estimate annual degradation rates (DR) and analyze the PV system behavior. The AI models, developed for each microinverter using historical operational data, showed high accuracy with normalized root mean square error (NRMSE) values between 1.64% and 2.92%. The DR, derived from the comparison between modeled and measured DC energy outputs, ranged from -2.91% to -12.85%, revealing a consistent decline after the third year of operation. These results were corroborated by both the effective peak power method and the I-V flash test measurements. Selected key performance indicators demonstrated seasonal variations and long-term degradation aligned with the estimated DR trend. The findings confirm the reliability of the AI-based approach and provide essential insights for predictive maintenance, inverter placement, and BIPV system optimization in urban settings.