Abstract: Considering the uncertainties of wind turbines output and the random fluctuations of load, and the correlation of multi-wind turbine output and load correlation, this paper first uses Latin hypercube sampling to generate multiple scenarios, and then uses K-means clustering to generate several typical scenarios. For K-means clustering, it is not possible to determine the optimal number of clusters based on the characteristics of wind power outpu t data and load data distribution. The clustering validity index is used to determine the optimal number of clusters. Then the non-dominated sorting genetic algorithm-II (NSGA-II) was used to solve the reactive power optimization model, and finally the simulation was performed in the improved IEEE33-bus power distribution network. The results prove that the reactive power optimization can effectively improve the voltage level of the distribution network and reduce the network loss.