Operation Optimization of Integrated Thermoelectric Energy System Based on Interruptible Load Demand

Aiming at the scheduling optimization problem of combined heat and power systems (CHPs) with demand response, a scheduling strategy based on deep reinforcement learning (DRL) is proposed. Firstly, the CHPs scheduling optimization problem is modeled as a Markov decision process (MDP); secondly, based on the economic scheduling of CHPs using the deep Q network (DQN) algorithm, through interruptible electric load compensation, the introduction of electricity demand response, further reduces energy consumption costs. Studies have shown that DRL can give full play to the autonomy of the model, and obtain optimal actions through continuous interaction with the environment; through interruptible load demand response, CHPs costs are reduced by 3.50%. Finally, the DQN algorithm is verified by the mixed integer linear programming model (MILP).