Wind Power Probability Density Forecasting Method Based On Data Mining And Non-linear Quantile Regression

Clean and renewable energy such as wind power is connected to the power grid on a large scale in order to solve the problems of traditional fossil energy shortage and environmental pollution.But it is more difficult to operate the power system safely and steadily.Wind energy is the fastest growing renewable energy source in recent years due to its wide distributed,never depleting,and the low cost.Affected by the intermittent and uncertainty of wind power,wind power prediction gradually has a key situation in the power system dispatching plan to solve the problem of wind power fluctuations.Energy Internet combines information technology and new energy technology support decision making through data integration analysis,which promotes energy sharing.Under the environment of Energy Internet,the accurate forecasting of wind power not only can reduce the loss of system spinning reserve capacity and conventional energy,but also can quantify the uncertainty and ensure the reasonable planning and scheduling of power system.Nowadays,the research on wind power forecast at home and abroad is mostly to obtain a deterministic wind power value or fluctuation interval,but cannot reflect the uncertainty of wind power.Probability density forecasting can reveal its volatility by predicting the possible occurrence probability of wind power at a certain interval in the future,and help the power sector to rationally plan the reserve capacity of conventional units and ensure the power balance in the system.In order to effectively quantify the uncertainty of wind power and improve the accuracy of prediction results,this paper proposes a probability density forecasting method based on the characteristics of wind power nonlinear time series.This paper constructs support vector quantile regression(SVQR)and neural network quantile regression(QRNN)model respectively,which combines support vector machine(SVM)and neural network(QR)with quantile regression(QR).At the same time,this paper combines the kernel density estimation with the above models to predict the probability density of wind power and obtains the accurate wind power range.In addition,in this paper,the outliers in the collected wind power data are considered.Thus it needs to adopt data mining technology to preprocess the raw data.The quartiles method and cubic spline interpolation are used to identify and correct outliers before short-term wind power forecasting.Then,a probability density forecasting method based on cubic spline interpolation support vector quantile regression(CSI-SVQR)is obtained.In order to measure the effectiveness and superiority of the above methods,this paper chooses mode,median and probability mean of the probability density curve as point predictions to make analysis,and the prediction intervals(PI)reliability criteria are used to evaluate the prediction intervals.This paper selects wind power historical data in Ontario,Canada,and Jilin,China as case study.The results show that on the basis of considering the nonlinearity of wind power time series,this paper can obtain the complete power probability density curves and accurate prediction intervals.It can better solve the problem of wind power uncertainty and provide technical support and theoretical basis for rational planning and dispatching power system.