Research On The Combined Forecast Model Of Wind Power

With the continuous development of wind power technology, wind power has taken bigger and bigger proportion in the power grid. The increment of wind power penetration threatens the security, stability, economic and reliable operation of the power system. Accurate prediction of wind power can reduce the operation cost and spinning reserve of power system and improve the wind power penetration limit, and it is conducive to the dispatching department to timely adjust the program, thereby to reduce the impact of the wind power on power grid. This paper studies two kinds of combination models in different forms, namely horizontal combination model and vertical combination model. The former includes the combination model based on different optimization criterions and the combined model based on induced ordered weighted harmonic averaging operator.The horizontal combined model based on different optimization criterions is proposed. The high accuracy single forecasting models are selected according to calculating the approach degree of every single model which can effectively solve the selection problem of single forecasting models. Taking use of three combined models with different optimization criterions to combinate, it can get the optimization model taking into account different optimization criterions. This can effectively overcome the problem that the combination model built according to some optimization criteria can not improve other evaluation indexes. By the actual example and simulation,The optimization model has a good overall evaluation index and can improve the forecast precision of wind power.The paper combines the maximum-minimum approach degree with induced ordered weighted harmonic averaging operator(IOWHA), and constructs the other horizontal combination model, namely IOWHA combination model.This model can conduct on the sequential assignment according to prediction accuracy level of each individual method in every moment, so it can effectively overcome the shortcomings of the traditional combined models. Using prediction value of the combined model with a high precision built by evey single forecasting model as standard to calculate the induced value of every single model, can solve the problem that the actual value in the forecast period are unknown, so the induced value can not be known in advance. The simulation results show that, the IOWHA combination model can finely reflect the tendecy of the change of future wind power and gain very high predicting accuracy.Study of wind power multi-step prediction method: based on the improvement of the horizontal combined model, the paper proposes the vertical combined model based on wavelet decomposition and phase space reconstruction technology. According to the characteristics of each component, they are divided into low frequency components and high frequency components; according to the characteristics of each frequency, select the appropriate single prediction method, the low frequency components reconstructed by means of phase space reconstruction technology are forecasted by Radical Basis Function(RBF) neural network; while the high frequency components are forecasted by rolling time series method; The final consequence can be obtained by combinating the forecast result of each component. The vertical combination model combines the powerful generalization ability and universal approximation ability of RBF neural network,with the advantage of updating the model according to data changes of rolling time series method. Simulation results show that the method has a good multi-step prediction ability and can realize multi-step prediction with high precision.