Optimal design of energy storage-supply systems using a multi-objective evolutionary algorithm and mixed-integer linear programming with a two-stage rolling horizon method

A near-optimal solution method for multi-objective optimal design problems of energy storage-supply systems is developed by hierarchically integrating a multi-objective evolutionary algorithm and mixed-integer linear programming with a two-stage rolling horizon method. The large-scale optimal design problem is decomposed into an upper-level design problem and a lower-level long-term operational planning problem. The design problem is solved using a multi-objective evolutionary algorithm (NSGA-Ⅱ) to effectively search diverse Pareto frontier. The long-term operational planning problem with the fixed design variables is solved by a two-stage rolling horizon method, in which energy storage planning obtained by solving the relaxed problem is used as a reference value and the short-term operational planning problems with tracking constraints are sequentially solved. To obtain the Pareto frontier, two-level optimization calculations are iteratively performed. The developed method is applied to a multi-objective optimal design problem of an energy storage-supply system including a photovoltaic panel, a water electrolyzer, a metal hydride tank, and pure hydrogen fuel cell cogeneration units. As the objective functions, the annual total cost and the annual exchanged electricity are minimized. The results show the high searching capability and computational performance of the developed method compared to a conventional solution method directly using commercial optimization solvers. • Near-optimal solution method based on NSGA-II with two-stage rolling horizon method. • Multi-objective optimal design considering year-round operational planning. • Decomposition into upper-level design and lower-level long-term operational problem. • Treating design weight as variable enhances diversity of Pareto-optimal solutions. • Higher computational and searching capability than other conventional methods.