The proportion of new energy in novel power system is gradually increasing, leading to a development trend of low carbonization, intelligence, diversification, safety and economy in novel power system, and an increasing demand for network security. With the rapid development of the national economy, the coupling dependence between the novel power system topology (physical layer) and communication network (information layer) is increasing day by day. Although this is conducive to the expansion of the power system scale, it is extremely prone to cascading failure faults, which can exacerbate faults and lead to large-scale power outages. On this basis, taking the cascading failure of the cyber physical system (CPS) in novel power system as the research object, a Markovbased dynamic coupling model for power cascading failure is constructed. The multi-level dynamic coupling relationship between power nodes and communication nodes is characterized, and the robustness of the model is analyzed. A key node identification algorithm for cascading failure of power CPS is proposed, and a path optimization model based on weighted composite index and influencing factors is proposed. The experimental results indicate that device safety plays a decisive role in optimizing the communication path for CPS cascading failures in novel power system, which can effectively save 0.52% of network energy consumption, with the lowest energy and optimal path.