Considering the load rate grade of power grid lines, and considering the unbalanced distribution of power flow changes based on Theil Entropy Theory, this paper proposes state importance index of nodes. In order to make the evaluation results more consistent with the complex network and the actual grid structure model, this paper proposes maximum flow contribution betweenness and closeness index. This paper combines the entropy weight method and the analytic hierarchy process to distribute the weight of the index , and obtains the comprehensive evaluation index of the key nodes, which not only takes into account the changes of the power grid state and the power grid structure, but also takes into account the subjective experience and objective data, so as to identify the key nodes of the power grid ac-curately. In this paper, the feasibility and practicability of the comprehensive index are verified by testing the IEEE-39 bus system.