Deep Reinforcement Learning of Cyber-Resilient Scheduling for Integrated Energy Systems

Integrated energy system (IES) plays an important role in promoting the consumption of renewable energy sources (RES) and improving energy efficiency due to its multi-energy synergy and efficient energy management. However, IES is highly dependent on information and communication technology, which exposes it to increasingly serious network security threats. Heat load redistribution attack (HLR), as a typical false data injection attack (FDIA), tampers with heat load scheduling signals, interferes with system decision-making, and seriously weakens the operational stability and economic performance of IES. In response to the threat posed by HLR attacks, this paper proposes a flexible scheduling optimization method based on the state-adversarial soft actor-critic (SA-SAC) algorithm to improve the anti-attack capability of IES in network attack scenarios. Firstly, a dynamic modeling framework of HLR attack is constructed to quantify the impact of the attack on thermal load scheduling and multi-energy network. Secondly, by incorporating adversarial learning into the training process, the algorithm enhances the robustness of scheduling decisions against malicious cyber intrusions. The simulation results show that the proposed method effectively mitigates the impact of HLR attacks and improves the robustness of system operation compared to traditional scheduling methods.