Intelligent hydrogen-ammonia combined energy storage system with deep reinforcement learning

Abstract

To achieve carbon neutrality, hydrogen and ammonia are considered promising energy carriers for renewable energy. Efficient use of these resources has become a critical research focus. Here we propose an intelligent hydrogen-ammonia combined energy storage system. To maximize net present value (NPV), deep reinforcement learning (DRL) is employed for the energy management strategy, dynamically adjusting the priority between hydrogen and ammonia. The results indicate that the DRL pathway achieves the highest NPV of 1.38 M$, which is 194 % of the benchmark pathway. Furthermore, the DRL pathway utilizes energy resources more efficiently, its grid dependency portion is lower than that of the benchmark pathway, particularly in November, by less than 0.8 %. Compared to conventional ways, the DRL pathway achieves zero carbon footprint, equivalently reducing 4819 tons, 17,715 tons and 94,944 tons of CO₂ emissions for ammonia, hydrogen and electricity production, respectively. Considering the carbon tax policy, this pathway could save up to 5.87 M$ annually.