Wind Power Forecasting Based On LWT And Lssvm Method

Sustainable development is advocated today, so how to use clean energy has become a prerequisite for social development. Among them, the wind power, which has rich resources as a clean energy, has been concerned widespread. However, the strong volatility and randomness of wind has brought great difficulties in the development. If the wind power can be predicted accurately, the operating costs of wind power systems can be reduced, in which way wind power will be made use of better and better.In order to find an effective way to predict the wind power, this paper will present Lifting Wavelet Transform(LWT), Support Vector Machine(SVM) and Error Forecasting(EF) methods in different stages of wind power forecasting process. LWT can not only play a de-noising effect, but also can extract the main variation of wind power; SVM is used to predict decomposition. The forecast accuracy and speed are guaranteed in this process; In order to reduce the large error points, EF method can effectively reduce forecast error and improve the stability of the results. A simulation experiment, which is based on the real data of wind farm in Inner Mongolia, confirms the effectiveness and feasibility of this method.In order to optimize prediction method, the method of Least Square Support Vector Machine(LSSVM) is used in training and prediction based on the above method. LSSVM uses the least square linear system as the loss function, instead of the traditional method of quadratic programming in SVM, which can simplify the computational complexity and improve the accuracy. Simulation results show that LSSVM is more effective to improve the prediction accuracy.Finally, wind power short-term load forecasting software is developed based on the above methods, using the GUI of MATLAB. The interface design is reasonable, the operation is simple, and the software contains a variety of forecasting methods. So, it can be operated for wind power forecasting.