The traditional network partition and pilot nodes selection methods for reactive power / voltage control are mainly based on the reactive power - voltage sensitivity matrix and clustering algorithm, however, it causes the imbalance of the reactive power and over-intensive pilot nodes in some partitions. According to the community structure partitioning algorithm in complex network theory, power system network modeling based on similarity weight is proposed, and an improved community modular index is put forward to improve the balance of the reactive power by introducing the reactive power balance degree of the network partitions. Based on the concept of node degree and betweenness centrality in complex network theory, a novel pilot node selection method is presented, which is based on the rank of observability and controllability sensitivity and can evaluate the centrality of node. Applying the proposed index and method to IEEE 39-bus system, simulation results show the method’s effectiveness.