Surrogate model-assisted optimization for multi-objective design of energy supply and storage systems considering one-year operational planning

A surrogate model-assisted optimization method is developed for the multi-objective design of energy supply and storage systems, considering one-year operational planning. The problem is decomposed into superordinate design and subordinate one-year operational planning problems. In the superordinate design, a multi-objective evolutionary algorithm (NSGA-II) is employed to search for diverse Pareto solutions. To improve computational efficiency, a radial basis function network surrogates to numerically solve mixed-integer linear programming-based one-year operational planning problems with searched design variables. The Pareto design solutions obtained by combining NSGA-II with the surrogate model are transmitted to the subordinate operational problems. The subordinate operational planning problems are solved using a two-stage rolling horizon approach, which can consider seasonal energy storage planning. To improve its prediction performance, the surrogate model is regularly updated using the one-year operational planning results additionally solved. These calculations are executed sequentially to search a near-optimal Pareto frontier. The developed method is applied to a system including a photovoltaic panel, a water electrolyzer, a pure hydrogen fuel cell, and a metal hydride tank. The objective functions are annual total cost and energy-supply autonomy. The results show 47% reductions in the calculation time and a better Pareto frontier compared to two-level optimization without the surrogate model.