Study On Ultra-short-term Probability Interval Prediction Of Wind Power Based On Run Theory

The burning of fossil fuel not only reduce the number year by year,but also increase the threat to the atmospheric environment.It is also becoming more and more urgent for people to explore,develop and utilize the new energy.As a clean energy source with abundant resources,wind energy has great development potential,but it poses a great threat to the stability of the power grid.Wind power prediction can not only provide reference for wind power plant's dispaching arrangement and maintenance,but also benefit the power system department to timely adjust the scheduling plan.The integrity of the data for wind power has the vital significance for the researcher's later work.We put forward a kind of missing data filling model based on the spatio-temporal correlation respectively form the time dimension and the spatial simension.First we put forward a kind of output weight optimization extreme learning machine,establish a goal through the point after the missing data.Consider the data before and after the missing data,comleting the missing data in the time dimension.For the spatial dimension,we use the granger causality test in this model,find the corresponding wind turbines,comlete the missing data in the spatial.Finally we use multiple imputation method,combinate the two groups of data.With the situation single or mult wind turbines losing data,we compare the results with ANFIS model,the filling effect is significantly improved.With the expansion of the wind power integration,high-precision wind power prediction is an important means to ensure the safe operation of wind power system.The large fluctuations and abnormal data may lead to the phenomenon of iterative decomposition of the classical EMD algorithm,so we introduce the weight function and modify the judgment condition of the mean value.Firstly,the wind power series is decomposed;Secondly,according to the characteristics of components,advance the run-test reconstruction to reconstruct components;Finally taking the wind power data of three farms in northeast China as example,using different methods on the real-time prediction.Comparing with other classical methods,the accuracy of the prediction result has been further improved.In this part,we analysis the error of the prediction result.Firstly,analysis the relationship between the prediction error of wind power and the capacity of wind power plant.Secondly,.we divide the wind power output into high output,middle output and low output.The probability density distribution of the prediction error is studied under different force levels.Finally combining with the nonparametric kernel density estimation,we propose the interval metood.Inthis paper,three indicators are presented to evaluate the interval prediction result.It is prove that the method is effective by comparing the parameterestimate method...