Node importance evaluation of electric power communication network is an important topic of power communication studies. For the problem of single connection weights and single evaluation index, the bandwidth and the distance of the electric power communication network are used for the network''s weight, computing the evaluating indicators of power communication network node, including joint strength, precision and node betweenness. Evaluation based on a variety of power communication network node and the fast density clustering method are used to establish the importance of the assessment model for power node communication network, supporting for the planning of grid communications. Using unsupervised classification by fast density clustering method, the node is divided into several fragile levels. This method can effectively improve the unsupervised classification method. The actual data grid communications are tested to verify the usefulness of this method in electric power communication network. This work is supported by National Natural Science Foundation of China (No. 61105115).