Research On The Wind Power Forecasting Method And Its Application

With the large scale wind power accessing power system,the volatility andrandomness of wind power brings serious challenges to the safety and stability ofpower system. Forecasting of the wind power is one of the key technologies to solvelarge scale wind power accessing system, which can provide support for powersystem dispatch control. At present, the forecasting method of wind power haveachieved some results,but not be researched in depth in China. The development offorecasting system is just started,lack of mature experience and standards. Therefore,the study of wind power forecasting methods is meaningful.To explore prediction approach of wind power with higher accuracy,the windspeed distribution,the power characteristics have been researched.Meanwhile, thewind power forecasting system is developed according to demand. The maincontents are as follows:Firstly,the wind speed and the power have been systematically studied in thispaper based on a wind farm. The result shows that the probability distribution ofwind speed presents Weibull distribution; wind power is random, fluctuations andchaos, with the time scale reducing the volatility of wind power is weaked.Secondly,a new forecasting approach for wind power based on Empirical ModeDecomposition(EDM) and phase space reconstruction is proposed.The decomposedIntrinsic Mode Function(IMF) component and the remaining components of thephase space reconstruct. The reconstruction of the sequence input Support VectorMachines (SVM) model to predict wind power. The results showed that after EMDdecomposition reduces the complexity of modeling, phase space reconstruction canreduce the impact of large deviations component for the predi cted results, whichimproved the accuracy of forecasting result.Thirdly,a nonlinear combination forecasting model based on meta-learning wasproposed.Meta knowledge formed by the results of Time Series Models, Grey Model,Linear Regression Model,Neural Network Model and feature attributes of series isused as input of meta predictor when combined forecasting is applied, system biascan be found and rectified.The weight of the base predictor are calculated usinggating network function in metal preditor.Weights of base predictors aretime-varying and non negative. The new algorithm is applied in wind power forecasting. Results show that the proposed method improves forecasting precisioncomparing with single forecasting algorithm and normal combined forecas tingalgorithm.Lastly,the wind power forecasting system was developed, which have beenachieved a short-term forecasting of wind power. The system is safty,easy to operateand comply prediction function exactly.