Accurate forecasting of wind power is an effective way to rationally adjust the scheduling strategies and to improve the operation stability and hence the economy of the power system with wind power or microgrids.A new improved prediction idea which does not rely on the basic methods is proposed based on analysis of a traditional wind power prediction procedure and of the strategy of how to use basic prediction methods in the wind power prediction.In the proposed method，an additional model is developed to calculate the prediction error，and the final result is acquired by adding the predicted error values to the traditional predicted values.Taking back propagation（BP）neural network as a basic prediction method，the proposed method is validated by predicting a real wind farm output.Simulation results show that the improved wind power prediction method can improve the prediction accuracy to a certain extent.The proposed method neither relates to the internal characteristics of the basic methods nor requires auxiliary method，which is different from traditional improved ideas and thus has a good versatility.