In order to improve the accuracy and speed of key node identification and damage resistance analysis of power grids, the topological structure and electrical characteristics are considered comprehensively. The topological information entropy in the improved K?shell algorithm (IKS) is replaced by power flow entropy, and the value obtained by weighted summing of the K-shell (Ks) value and power flow entropy is defined as a new index for evaluating node importance, namely the comprehensive value of node importance. The comprehensive value of node importance is ranked, and the key node identification model with IKS and power flow entropy is obtained. By taking the IEEE-118 node system as an example, the proposed method and four traditional key node identification methods are used to identify nodes, and random attack and deliberate attack tests are carried out with different attack intensities. According to the node importance ranking, the maximum connectivity subgraph scale of power grids, and network efficiency analysis, the proposed key node identification method based on IKS and power flow entropy is superior to betweenness centrality and closeness centrality.