A combination prediction model modeling method for wind power average value prediction and wind power fluctuation range prediction is proposed, which is based on the fuzzy information granulation and least squares support vector machine (LSSVM). Firstly, fuzzy information granulation of the training samples is made, and effective component information of each window is extracted according to the need, namely the minimum, average and maximum value of each window. Secondly, LSSVM of the prediction models are established for each component, and then the adaptive particle swarm algorithm is used to optimize each component model. Finally, the optimized LSSVM model is used for combined forecast in terms of wind power average value and wind power fluctuation range. The case study shows that the combined prediction model can effectviely predict wind power average value prediction and wind power fluctuation range, and accurately track the wind electric power change.