Volume 32: A Sustainable, Clean and Carbon-free Energy Future

Deep Reinforcement Learning Based Energy Scheduling of a Hybrid Electricity/Heat/Hydrogen Energy System Xianlian Wang, Junyi Dong, Li Sun

Abstract

Hybrid electricity/heat/hydrogen energy system is a potential solution for the future low-carbon residential energy system. This paper studies the efficient energy scheduling of such system, including hydrogen production, utilization and storage processes. To overcome the problems of coupling among the multi-energy flow and the uncertainties on both sides of power and load, a deep reinforcement learning (DRL) algorithm, namely deep deterministic policy gradient (DDPG), is used to realize adaptive energy scheduling of the system. The scheduling results of simulation experiment under typical winter day scenario illustrate that, based on the pre-trained DDPG framework, the system can achieve a rapid response to the environment and optimize energy efficiency. Additionally, by appropriate power charging and discharging, the states of energy storage devices can essentially recover to their initial states, enabling the sustainable operation of the hybrid energy system.

Keywords hydrogen fuel cell, DDPG, energy scheduling, hybrid energy system