Research On Short-Term Prediction Method For Wind Farm Generating Power

With the high-speed development of wind power and the permeability increases sharply in the grid,it makes the high randomness and volatility of wind power has more and more influence to power grid.Accurate wind power prediction is of great importance to the safe and economic operation,reasonable power generation scheduling,peak or frequency regulation of power grid.There have been a lot of problems with the studies that have been done,and this paper makes a deeper discussion on it.The traditional least squares support vector machine(LS-SVM)wind power forecast model only optimizes the model parameters(punishment factor ? and width of kernel function ?),and it does not excavate sample information to a large extent and explore the interdependence among them,so the forecast error is relatively large.In this paper,the space reconstruction parameters of wind speed time series(embedded dimension m and delay time ?)are included in the parameter optimization,the particle swarm optimization algorithm(PSO)is used to optimize four parameters ?,?,m,and ? at the same time.The case analysis of a wind farm in northwest China shows that the four-parameter simultaneous optimization method proposed in this paper is practical and feasible,and compared with the traditional two-parameter optimization forecast model,the forecast accuracy of wind power can be improved to a certain extent.The single short-term wind power point prediction often can not meet the needs of power grid risk assessment and decision-making.In this paper,we first calculate the theoretical probability model of each wind power forecast box by empirical cumulative distribution function(ECDF)in MATLAB,and then use the exponential covariance function expression to determine the best covariance matrix corresponding to the dynamic scenarios,and determine the multivariate normal distribution model of wind farm output obedience at multiple connected moments;For each predicted moment of the wind power point prediction value of the wind belongs to the prediction box,we direct sample random vector which obey multivariate normal distribution to form the wind power dynamic scenario.After a simulation experiment on a real wind farm,the results show that the scenario set considering wind power fluctuation at different time scales can cover the measured wind power curve and the reliability of the method is proved.In view of the limitations of Auto Regressive and Moving Average(ARMA)wind power prediction methods to analyze large error fluctuations,ARMA model is obtained based on historical data,and then the generalized autoregressive conditional heteroskedasticity(GARCH)model was used to eliminate the conditional heteroskedasticity of the wind power forecast error,thus an ARMA-GARCH composite forecast model was formed;and further based on the statistical characteristics of the forecast error peak and light tails,using improved generalized error distribution(GED)model that has advantages over other probability density distributions to propose the view of stratification for wind power forecast error,and gives corresponding compensation schemes for forecast error under different conditions.Finally,it is proved the effectiveness by example calculation that the proposed scheme improves wind power forecast accuracy.