Two-layer scheduling optimization for hydrogen-coupled integrated energy system based on model predictive control and deep reinforcement learning

To address current energy trends, integrated energy systems must achieve multi-energy complementarity, enabling oil, coal, natural gas, and electricity to interact synergistically within the system. Hydrogen, as a clean secondary energy, features high capacity, high energy density, and low pollution. When applied to IES, it plays a crucial role in enhancing the coupling between different energy forms. This is essential for improving system energy efficiency, operational flexibility, and reducing low-carbon emissions. This paper proposes a hydrogen-coupled integrated energy system (HIES) and designs a two-layer scheduling optimization framework based on model predictive control (MPC) and deep reinforcement learning (DRL). The upper layer employs a long short-term memory (LSTM) algorithm for load forecasting at a 15-min time scale for typical days. The lower layer utilizes deep deterministic policy gradient (DDPG), incorporating time-of-use electricity pricing and carbon pricing mechanisms, to perform low-carbon economic scheduling at 15-min intervals, forming a two-layer rolling optimization framework.