Optimal Operation of Integrated Energy System Considering Carbon Emission: A Case Study with Deep Reinforcement Learning

In this paper, the optimal operation of an integrated energy system consisting of renewable energy, gas boiler, combined heat and power plant, HVAC (heating, ventilation and air conditioning) system, battery storage, and thermal storage is investigated. This paper considers both system energy cost and carbon emission. To handle the challenge of system uncertainty from renewable energy, outside temperature, and electrical and thermal base load, a deep reinforcement learning (DRL)-based model-free approach is proposed. To verify the effectiveness of the proposed approach, a numerical experiment on real data is conducted. Experiment results illustrate that the proposed approach can save cost while maintaining system thermal comfort by benchmarking against model predictive control.