Reactive power optimization is an important measure to ensure the reliable operation of the system. According to the characteristics of reactive power optimization,a reactive power optimization method based on localized voltage stability index partition and improved particle swarm optimization algorithm is proposed. Firstly, the local voltage stability index of the system load node is calculated, and the load nodes are sorted according to the voltage stability index. A part of the load nodes are sorted as the candidate compensation points, and the objects are partitioned by electrical distance. Then, the improved particle swarm algorithm System optimal compensation point position and reactive power compensation amount;Finally, in MATLAB with IEEE33 node system simulation verification,the simulation results show that the method of combining the local voltage stability index with the electrical distance can narrow the optimization range, and obtain the candidate compensation area reasonable and effective. The improved particle swarm algorithm to initialize particle diversity better, with faster convergence speed.