Typhoons can have a serious impact on a distribution network, so it is very important to formulate the windproof reinforcement planning scheme of overhead lines. To improve the resilience to natural disasters, a method for improving distribution network planning based on the whole process simulation of typhoon disasters is proposed. By simulating the real-time wind condition of the whole process from typhoon landing to typhoon extinction, differentiated reinforcement is implemented for each line. First, the Dirichlet process mixture model (DPMM) clustering algorithm is used to extract typical typhoon landing scenarios, and the storm track model and Batts wind field model are combined to simulate the typhoon moving path and wind field, and the real-time fault probability of distribution network lines is calculated. Secondly, combined with the typhoon simulation results and the differential overhead line failure rate in different wind speed standards, a double-layer stochastic programming model is established with the objective of minimizing the annual investment cost of multi-level line reinforcement, the cost of loss of load during typhoon transit, the loss of power outage and maintenance cost. It is analyzed using Benders decomposition algorithm. Finally, taking the improved IEEE33 node system as an example, the effectiveness of the proposed method is verified.