In consideration of distribution networks with embedded wind power， a stochastic model of wind power generation and load is established. Reasonable reconstruction results can be obtained by adopting the proposed algorithm to reconfigure the network， which is called minimum spanning tree led by particle swarm optimization， using probabilistic power flow to analyze the impact of the random output of wind power and taking static unsafe probability of network and mathematical expectation of the network loss to evaluate the advantages and disadvantages of the reconstruction results. The proposed reconstruction algorithm combines the advantages of particle swarm algorithm and minimum spanning tree algorithm. It will avoid the invalid solution by particle swarm algorithm and guide building a spanning tree to be optimized. The proposed method applied for reconfiguration of distribution network is evaluated. The results show that the method can not only reduce the system power loss， but also effectively solve the impact of stochastic wind power on system stability.