Short-term Wind Power Forecasting Based On Frequency Domain Decomposition

As clean energy, the reserve of wind power is huge. But therandomness and the uncertainty of the wind power increase the difficultyof large-scale wind power integration and hinder the development ofwind power. If the wind power can be forecasted accurately, we can dealwith the wind power integration difficulties and promote the rapiddevelopment of wind power.In order to improve the accuracy of wind power load forecasting, themethod of frequency domain decomposition will be used in wind powerload forecasting in this thesis. This method can find the wind power lawand overcome the irregularity of wind power to a certain extent. With themethod of frequency domain decomposition, the original load data willbe decomposed into five parts, i.e. daily cycle, week cycle, month cycle,the low and high frequency part. According to the respectivecharacteristics of the five parts, the daily cycle part and high frequencypart can use the method of BP neural network to train and forecastrespectively; the low frequency part is a smooth curve, it can beforecasted through the method of a linear regression. The actual data areused in simulation, the results show the validity of the method above.In order to further improve the forecasting accuracy, we candecompose the daily cycle and the high frequency part with the threelayer lifting wavelet to remove the burr. And then, it can be trained and predicted using the method of the Least Squares Support Vector Machine(LSSVM). The low frequency part still uses a linear regression method toforecast. Finally, add the forecasting results together, the high precision ofwind power forecasting can be realized. The experimental results showthat the method based on the frequency domain decomposition, can findthe wind power law more easily, benefit for short term load forecasting bycombining the three methods for different parts, and greatly improve theaccuracy of forecasting.Finally, a short-term wind power forecasting system software isdeveloped based on the method above in MATLAB GUI platform. Thesystem has good compatibility and friendly interfaces. It can be operatedconveniently and forecast accurately.