Renewable-energy-driven integrated energy systems: Sensitivity–interaction analysis based on multi-objective mixed-integer linear programming optimization and surrogate modeling

Driven by the rapid growth of renewable energy and increasing demand for multi-energy complementarity, this study proposes an integrated energy system framework with a three-objective ε-constraint optimization model to characterize trade-offs among economic cost, primary energy consumption, and system operational independence. The obtained Pareto frontier reveals clear extremes of system planning strategies: the cost-oriented solution (458,570.28 $, 5,409,966.45 kWh, 34,439.56 kWh) minimizes investment but exhibits strong dependence on external energy, whereas the independence-oriented solution (1,040,608.05 $, 733,973.74 kWh, 3160.20 kWh) significantly enhances self-sufficiency at the expense of substantially higher cost, and the compromise solution achieves balanced performance with an annualized cost of 765,585.52 $, primary-energy indicator of 1,415,889.34 kWh, and grid-exchange level of 7902.53 kWh. In addition, sensitivity analyses of electricity price and electrolyzer (ELC) efficiency, TOPSIS-based weight sensitivity, and multi-scenario load uncertainty analysis are conducted. It is found that rising electricity prices promote energy efficiency and system independence but impair economic performance, while improved ELC efficiency enhances energy utilization and autonomy with relatively stable costs. The TOPSIS decision results show strong robustness against moderate weight perturbations. Meanwhile, increased load levels degrade system economics, energy efficiency, and operational independence, presenting intensified supply–demand balancing challenges. To further uncover the underlying driving mechanisms of these trade-offs, a surrogate-assisted global sensitivity analysis framework is embedded into the optimization process, integrating permutation–SHAP importance evaluation, second-order interaction attribution, and Sobol indices. The results consistently identify wind power capacity, hydrogen storage, proton exchange membrane fuel cells, and ELC as dominant factors influencing system performance, while revealing strong interactions among key component pairs, such as wind–battery, fuel cells–hydrogen storage, and absorption chiller–battery.