Deep Reinforcement Learning Based Approach for Real-Time Dispatch of Integrated Energy System with Hydrogen Energy Utilization

Abstract

Integrated energy system (IES) with multi-energy coupling can improve energy utilization efficiency and reduce carbon emissions, and therefore have received widespread attention. With a large number of renewable energy sources and multi-energy loads connected to the integrated energy system, the uncertainty at the supply side and demand side poses a great challenge to the optimal dispatch of IES. To cope with the gap, a real-time optimal dispatch method based on deep deterministic policy gradient (DDPG) is proposed to solve the optimal dispatch of IES considering hydrogen energy utilization. The mathematical model of the problem is established and modeled as a Markov decision process (MDP). Based on this, a deep reinforcement learning (DRL) framework is established and the DDPG algorithm is used to train the agent offline. The trained agent enables online real-time dispatch decisions. The proposed method has advantages in terms of operating cost and decision time. In addition, the advantages brought by the hydrogen utilization device for system operation are verified.