A long-term field degradation and reliability assessment of CdTe BIPV systems using AI-based performance analysis

This study assesses the long-term performance degradation of a Building-Integrated Photovoltaic (BIPV) system using cadmium telluride (CdTe) modules over six years of field operation. A methodology based on Gradient Boosting Regressor models was used to estimate AI-based cumulative degradation loss ( L AI ) with uncertainty intervals, while effective peak power analysis and laboratory I-V curve measurements were used as complementary validation methods. The AI models achieved high predictive accuracy, with normalized root-mean-square errors ranging from 1.64% to 2.92%. The calculated L AI increased from 2.91% in 2019 to 12.85% in 2024, with most losses occurring during the first three years before stabilization. The effective peak power method indicated annual degradation rates of approximately −1.7% to −1.9%/year, while I-V measurements showed a total power loss of about 10.2%. The agreement among the three methods supports the observed degradation trend. In addition, selected key performance indicators exhibited seasonal variability and long-term performance decline, consistent with the estimated degradation behavior. The results highlight the value of long-term monitoring and AI-based analysis for degradation assessment, predictive maintenance, and improved reliability of CdTe BIPV systems.