Research On Short-term Prediction Of Wind Speed

The wind speed fluctuation will cause the output fluctuation of the wind farm and the frequency fluctuation of power system and will impact the power system safty and stability. Accurate wind speed forecasting may relieve the wind farm output fluctuation, helps electric power system dispatchors make dispatching schedule in advance and has important significance for the economic integration of wind power.Firstly, Back-propagation (BP) neural network model for predicting wind speed is suggested. Three-layer BP neural network was employed for short-term wind speed prediction. From the number of input neurons, different training functions and the number of hidden layer neurons perspective, the predicted results were compared and analyzed.Secondly, by using wavelet analysis method the wind series were decomposed into different wind speed sub-series, and then the sub-series were predicted by BP neural network model respectively. The predicted results were added as the final predicted wind speed. The prediction results were compared and analyzed under the conditions of the different parameters.Finally, wavelet packet analysis method was used for the prediction of wind speed. Compared with the wavelet analysis method, the wavelet packet analysis method increases the decomposition of high frequency components. At last, the three prediction modles were compared and analyzed.