A hybrid ML-MILP framework for the optimal integration of photovoltaic and battery systems in manufacturing industries

Abstract

The increasing integration of renewable energy sources, coupled with volatile electricity prices, poses significant challenges on energy-intensive industries seeking to reduce costs and improve energy efficiency. This work presents a novel hybrid framework combining machine learning (ML) predictive algorithms with a mixed-integer linear programming (MILP) model to optimize energy management in manufacturing industries utilizing photovoltaic (PV) systems and battery energy storage systems (BESS). The proposed framework accurately forecasts electricity prices, PV generation, and industrial energy demand, enabling both operational optimization and strategic investment planning. The MILP model ensures efficient energy resource utilization, by minimizing electricity costs and maximizing financial gains through optimal market participation. The framework was validated through a real-life case study of a Greek manufacturing facility, comparing different energy options, including scenarios with and without BESS. Results revealed that a properly sized BESS can significantly facilitate cost savings of up to 352 RMU¹/day via price arbitrage, especially during peak pricing periods. Further analysis indicated that increasing BESS capacity could yield even higher financial benefits thus enhancing industry profitability and competitiveness. Sensitivity analysis under varying electricity price scenarios confirmed the robustness and adaptability of the proposed framework to dynamic market conditions. Financial analysis highlighted that, with appropriate subsidies, the payback period for BESS investments could be considerably shortened from 7 years to 4 years, improving feasibility and attractiveness.