As the wind power penetration is increasing ceaselessly, the influence of wind power for voltage stability should not be ignored. According to the characteristics of wind power, this paper proposes an improved continuation power flow method. This method considers the wind generators switching to power-reducing mode when the voltage declines. At the same time, this method considers using the wind farm’s remaining capacity of reactive power compensation device to improve the condition of grid voltage. To study the influence of load uncertainty on system, the probabilistic load flow is used to make voltage stability analysis. The improved continuation power flow method is used to research the samples of probabilistic load flow. At the same time, the fuzzy C-mean clustering is used to find the load bus group. A Latin hypercube probabilistic load flow model based on load clustering is built. The validity of proposed method is tested on IEEE 14 and IEEE 57 bus system, which have wind farms. The simulation result shows that the method of this paper matches the practice better. Besides, this method can improve the sampling efficiency. This work is supported by Key Research and Development Program of Shanxi Province (No. 201603D112001).