Research On The Combination Method Of Forecasting The Ultra-short-term Power Of Wind Farm

Wind power generation is affected by wind energy characteristics.Due to the randomness of wind energy,wind power has uncontrollable characteristics such as intermittent and random volatility.With the development and utilization of a large range of wind energy,wind power grid capacity is increasing.The potential risk of wind power fluctuation is increasingly obvious.The ultra-short-term prediction of the output power of wind farms is one of the effective techniques to solve the problem of power network dispatching control.Because the accuracy and stability of traditional single prediction method can't meet the application requirements,so combining the advantages and characteristics of a variety of forecasting methods to carry out the combination forecast,which is an effective means to improve the accuracy and stability of the ultra-short-term prediction of wind power.Based on the measured data of a certain wind farm in China,this paper establishes a combined forecasting model studies the ultra-short-term forecast of wind power,which by combining Theil unequal coefficient and the improved induced order weighting operator.At the same time,an integrated wind power forecasting platform is designed according to the demand.The main contents of this paper are as follows:Firstly,the data preprocessing method is given for the wind farm parameters,and the characteristics of the wind speed and wind power parameters are analyzed.At the same time,the sources of prediction error are studied,and the evaluation system of multi-index prediction error is established.Secondly,Wavelet Neural Network(WNN),Genetic Algorithm to optimize the BP Neural Network(GA-BP),Support Vector Machine(SVM)and the method of time series named Auto Regressive Moving Average model(ARMA)models were used to doing ultra-short-term prediction with 10 min and 1 hour time scales.The results show that SVM has the best predictive effect at 10 min ahead of time,but for 1h ahead of schedule,GA-BP is the best.On the basis of the single prediction models,this paper proposes a combined forecasting algorithm based on the Theil inequality coefficient and the improved induced ordered weighted operator.The error information matrix is used to analyze the redundancy,and found that the WNN model is a redundant model.Since the actual value of the predicted time is not known in practice,the three induced ordered weighting operators can't be used directly.So they are improved,that is,the prediction precision mean of the first few moments of each single prediction model replaces the induced value of three operator.By analyzing the time scale of 10 min and 1h ahead of time,the results show that the IOWA combined model can effectively improve the short-term prediction accuracy of wind farm output power.Lastly,a set of integrated wind power forecasting platform is designed.The platform design scheme is introduced in detail through the software function module,the platform communication design and the platform client design,and some of the research results were demonstrated.