In light of the digital transformation of power systems and the increasing integration of renewable energy, grid control architecture has gradually evolved into deeply coupled cyber-physical power systems (CPPS) with chain-like and coordinated security risks. A Digital Twin-Based Framework for Attack Chain Modelling and Defence Strategy Verification in Power Grids is proposed in this paper. A bidirectional mapping model of the physical and cyber layers is built in this way to detect abnormal behaviour in time and space, and then adjust. A data-driven digital twin model of LSTM networks is used to address typical denial-of-service (DoS) attack scenarios and realize attack detection, alert, and self-healing decision verification. According to the results of the simulation, the proposed method can effectively reconstruct the process of the attack chain and improve the robustness and adaptive recovery capabilities of the control system in multi-node parallel systems significantly.
