Hybrid Regression–Artificial Bee Colony Optimization for PV Production Forecasting under Energy Performance Contracting

Accurate photovoltaic (PV) energy production forecasting is essential for Energy Performance Contracts (EPCs), where financial outcomes and contractual guarantees depend on reliable performance estimates. This study proposes a hybrid forecasting framework that integrates multivariate regression analysis with the Artificial Bee Colony (ABC) algorithm to improve prediction accuracy while preserving computational efficiency and model transparency. The proposed model is validated using real operational data from a 1710.72 kWp grid-connected PV system operating under an EPC framework at Alanya Alaaddin Keykubat University (Türkiye). Key technical and economic variables, including solar irradiance, investment cost, and electricity unit price, are employed in the regression model, whose coefficients are optimized using the ABC algorithm. Results show that the hybrid Regression–ABC model achieves a MAPE of 6.82%, significantly outperforming the baseline regression model (14.67%). The predicted annual energy production closely matches measured field data, with a relative deviation of approximately 0.01%, remaining within typical measurement uncertainty. The findings demonstrate that the proposed hybrid approach provides an accurate, transparent, and practical forecasting tool suitable for EPC-based PV projects, supporting performance verification, risk management, and investment planning.