Abstract: Solving the poor constraint of combining meteorological factors data with power data and the decay of historical time series information of wind power in traditional wind power prediction,a wind power prediction method based on multiple joint probabilities and improved weighted Hidden Markov Model（HMM）is proposed in this paper. Meteorological factors in Numerical Weather Prediction（NWP）are combined by multiple joint probability firstly. Subsequently,the NWP data and the power time series are fused by improving the release probabilities in the HMM to constrain each other. Then,the multi-step predicted values are weighted by the conditional entropy improved rough set with improved conditional entropy,and obtain the final wind power prediction results. Finally,the wind power prediction accuracy can be effectively improved by fusing and confining NWP data and power data with each other,as verified by actual arithmetic cases in wind farms.