Under the global context of climate change and energy transition, the structural transformation of energy systems in industrial parks is recognized as critically imperative. However, traditional small and medium-sized industrial parks in China are confronted with challenges including the lack of theoretical models and operational validation during green power substitution. To address these issues, a multi-objective optimization framework is developed in this study, integrating an enhanced NSGAII algorithm with an adaptive real-coded simulated binary crossover (ARSBX) operator and a source-load-storage collaborative optimization strategy. The integrated energy system of industrial parks is optimized through green power substitution, thereby revealing nonlinear coupling mechanisms between energy consumption and economic costs before and after the transition. Three steps are outlined: First, constructing an integrated energy system using the NSGAII-ARSBX optimization algorithm to facilitate green power transition. Second, establishing a source-load-storage coordinated optimization model by integrating a random forest algorithm and mixed-integer linear programming to minimize system costs. Third, validating the methodology through an empirical case study in a Chinese industrial park. Results demonstrate that compared to conventional energy supply systems, the proposed methodology achieves a 78.81% improvement in renewable energy utilization, a 41.93% reduction in energy costs, and a 78.81% decrease in carbon emissions. This research is demonstrated to provide theoretical and practical significance for advancing green power substitution and low-carbon transitions in traditional industrial parks.