Capacity planning for integrated energy system based on reinforcement learning and multi-criteria evaluation

Optimal capacity planning for energy devices is significantly crucial for saving economic costs and enhancing operational efficiency in an integrated energy system (IES). In this study, a reinforcement learning (RL)-based capacity planning approach for IES is proposed, where a multistage decision-making strategy is designed to reduce the action dimensionality for improving computational efficiency. Besides, to evaluate each capacity configuration scheme in RL process, a sound multi-criteria system from aspects of economy, environment, efficiency and safety is built. Within it, some new indicators are innovatively developed, including a capacity factor criterion, an installed capacity ratio and generation ratio of renewable energy devices as well as an adjust rate of energy generators. To verify the effectiveness of the proposed approach, a case study for an industrial park is carried out. The experimental results demonstrate that the proposed approach outperforms the conventional mixed integer linear programming (MILP) and multi-objective optimization (MOO) -based methods on planning the optimal capacity. Furthermore, comparative experiments with separate production systems are conducted to further study the advantages of IES, and sensitive analyses are also performed to verify the robustness of determining the optimal capacity configuration of this proposed method.