Optimizing Solar Photovoltaic Module Supply Chains through Artificial Intelligence-Driven Mathematical Modeling

Abstract As solar photovoltaic installations increase rapidly, the demand for optimized supply chains becomes critical. However, most previous photovoltaic module supply chain studies rely on fixed parameters and rarely integrate machine learning-driven parameter estimation into system-wide optimization frameworks. Therefore, this paper proposes a hybrid framework that integrates machine learning with mixed-integer linear programming mathematical modeling to improve the efficiency of the photovoltaic module supply chain and reduce its costs. Critical parameters, including production quantity, production costs, energy usage, and transportation, were forecasted with machine learning models, such as CatBoost, LightGBM, and Random Forest, trained on datasets derived from industry reports and scholarly articles. Two parallel models were examined: a baseline model with constant parameter values and a machine learning-enhanced model that employs estimated inputs. The results showed that the machine learning-enhanced model reduced total photovoltaic module supply chain cost by 9.50%, achieving savings of approximately $1.34 billion in the system examined in this paper. Sensitivity analysis identified production cost and demand level as the dominant cost components. The main contributions of this study lie in the integration of machine learning-based parameter prediction with mixed-integer linear programming optimization, the quantitative comparison between machine learning-enhanced and baseline photovoltaic module supply chain models, and demonstrating the potential of artificial intelligence-driven optimization to improve forecasting accuracy, resource allocation, and operational efficiency in renewable energy supply chains.