Research On Wind Power Forecast Based On Grey Theory

As global energy problem becomes increasingly serious,the research and utilization of renewable energy has become a hotspot in society.Wind is regarded as the most potential development and most widely use of green energy,with the advantages of safety,non-polluting,renewable,environmental protection and ecological benefits friendly.Accurate wind power prediction is helpful to solve the problems of wind power output power control,power grid safe and economic dispatch and wind power bidding trading,but wind power series has a high degree of non-stationary and uncertainty,the accurate forecast is still need further research.This dissertation discusses short-term wind power forecasting in wind farm,and researches on its performance improvement.In this paper,wind power series were decomposed by wavelet decomposition to reduce the non-stationary,and then augmented Dickey-Fuller(ADF)test was adopted as a test method for the stationary of each decomposed component after wavelet decomposition.Furthermore,since it is difficult to forecast the wavelet detail characterized with high frequency because of its fluctuation,based on grey theory,a second order gray forecasting model is presented to forecast each component after wavelet decomposition.The wind power interval prediction model is obtained by analyzing the error transfer series of the fitted values with grey models and establishing Markov state transition probability matrices,to overcome the uncertainty of wind power series.Major work and contributions of this dissertation include:(1)First,the background and significance of wind power forecasting are expounded,the current development status of wind power forecasting technology both in China and abroad.According to the existing research results,the research work of this paper is arranged.(2)This paper introduced the characteristics and basic principles of wavelet transform.And then augmented Dickey-Fuller(ADF)test was adopted as a test method for the stationary of each decomposed component after wavelet decomposition.The test results provide the basis for the best wavelet decomposition level.(3)Furthermore,since it is difficult to forecast the wavelet detail characterizedwith high frequency because of its fluctuation,based on grey theory,a second order gray forecasting model is presented to forecast each component after wavelet decomposition in this paper.In order to obtain the optimum parameters of second order gray forecasting model,solving the parameters of grey model as initial value by least square estimation,and neural network mapping approach is used to build the second order gray neural network forecasting model(GNNM(2,1)).The simulation result verified that the method effectively improved the wind power forecasting accuracy.(4)In order to obtain more information about wind power,a wind power interval prediction model based on wavelet decomposition second order gray neural network model and Markov chain is proposed in this paper.The wavelet decomposition second order gray neural network model is presented to forecast wind power point,the wind power interval prediction model is obtained by analyzing the error transfer series of the fitted values with grey models and establishing Markov state transition probability matrices.Firstly,K-means clustering algorithm is employed to solve the problem of the state division of Markov Chain.Markov chain is applied to study the transfer characteristics of absolute value of prediction error.The expected values of the transfer matrices are used to the correct the tradition Markov chain.The simulation result verified that the method effectively improved the wind power forecasting accuracy.(5)Finally,the research work of this paper is summarized and analyzed,further noted their application scope and development direction.