This paper discusses the reactive power optimization problem in distribution network with wind power integration and presents a new algorithm for it. In view of the uncertainty feature of wind power, mathematical model is built by the scenario analysis method and the improved particle swarm optimization algorithm (IPSO). Based on the Wasserstein distance, the optimal scenarios for wind power and the corresponding occurrence probability are deduced. By using the optimal scenario analysis results, the objective function is established according to the power loss and voltage stability margin. The optimization procedure is processed by the improved particle swarm optimization algorithm. The crossover and the mutation operator are added to the IPSO algorithm. In the beginning, classical PSO algorithm is carried out, and after 20 iterations, the crossover and the mutation operator begin to play a role. To ensure the diversity of particles, the mutation probability (Pm) is improved, and it can self-adjust the mutation operator based on the degree of convergence. By this way, the IPSO can both maintain rapid convergence speed and better global astringency. The IEEE 69-bus distribution system is used to the experiment, and test results show that the model and the algorithm are effective.