Study On Optimal Scheduling Model Considering Wind Power Prediction Error

The global shortage of ore energy has brought great inconvenience to human life,but it has also become an opportunity for the prosperity and development of green new energy.With the growing maturity of related technology,wind power has become the leader in the development of new energy industry,which combines the advantages of rich resources reserves and clean pollution-free of renewable energy.However,the inherent randomness of wind energy results in a large fluctuation in wind power output and it is difficult to predict accurately.Dispatch decision making is very difficult,as a result the wind power has a lot of waste,the advantages of wind power generation can not be full play,and the system economy is great decline.Aiming at the problem of the difficulty of frequently changing wind power consumptive in large power grid,and analyzing the characteristics of wind power output,and then developing the coordinated optimal dispatching model,which through the form of statistical information of forecasting error reflects the random variation of wind power output and considering the high energy load with electric self-adjusting ability participate in the dispatch to deal with the change of wind power output.The main research work is outlined below:Through researching the characteristics of changes of wind power output in different lengths of time,the frequently fluctuating and rather undulating characteristic of wind power output is clearly defined.The change characteristic is reflected in the form of forecasting error.The distribution characteristics of the prediction error are calculated from the power side,the month side and the time series side,and then analyzing the difference of the error distribution.In order to fully describe the characteristics of the error,the error is segmented on these three sides.The number of segments obtained by rough segmentation is excess,the distribution of data in each section is uneven,and the result in the error description is easily trapped in the details.In this thesis,a cluster analysis method is proposed to merge the segments with larger similarity of error characteristics for taking into account the whole feature of error distribution.To take full information of error for reducing the difficulty of scheduling decision,this thesis studies the probability density fitting of errors ba sed on the data partitioning.The normal distribution,the t-location-scale distribution and the nonparametric kernel density estimation method are applied to the error fitting.And then the efficiency coefficient and the consistency coefficient that were widely applied in the hydrology filed are introduced to compare the performance of each fitting model.In the use of kernel density estimation,the method of window width determination which is easy to be overlooked in previous research is concerned,therefore an iterative method is proposed to select window width.All evaluation of the fitting of error data for each group reveal the robustness of non-parametric method for the pinnacle and trailing distribution,and the iterative method also shows superior performance.The error interval estimation is carried out on the basis of non-parametric fitting,which is a groundwork for bring the uncertainty of wind power into the subsequent dispatching model.By analyzing the power consumption characteristics of hig h load energy,it is determined that the discretely regulative high energy load and the self-generation power plant which can serve as virtual high energy load participate in the power grid dispatching.For reducing costs and improving wind power consumpti on,the multi-objective optimization model is established.In order to take account of the uncertainty of wind power in the optimization schedule,the error information is introduced in the form of prediction error confidence interval in the economic objective function.An improved bat algorithm based on decomposition is used to solve the built model.The correctness and feasibility of the established dispatching model are verified by an example analysis.