Research On Generation Method And Application Of Typical Scene Of Wind Power

Wind power is a kind of clean energy with abundant reserves,which accounts for a large proportion of the energy structure.However,wind power has significant randomness,volatility,and peak inversion characteristics.When large-scale wind power is connected to the grid,the existing conventional thermal power units are forced to participate in the system peak shaving,which increases the difficulty and cost of grid dispatching.Secondly,due to the limitations of the grid structure and load level,the grid’s ability to accept wind power is limited,resulting in the occurrence of "abandonment of wind".Power system scenario analysis is widely used to describe the uncertain factors of wind power output,but too many scenarios will increase the complexity of power system planning and scheduling.For the above problems,on the one hand,it is necessary to model the uncertainty of wind power output to improve the accuracy of wind power output prediction,and to solve the uncertainty problem into a deterministic problem through scene analysis methods;on the other hand,in the power Increase the flexibility of power supply in the system’s recent scheduling,improve the system’s peak shaving capacity and wind power acceptance capacity.In this paper,the current probabilistic modeling of wind power needs to assume a parameter distribution,which cannot fully consider the impact of various random factors.A non-parametric kernel density estimation theory is proposed to carry out probabilistic modeling of wind power,and an optimal bandwidth improvement is given.Method,through the goodness-of-fit test to test the fitting effect of the kernel density estimate.Secondly,in terms of scene generation,this paper focuses on a wind power scene generation method based on unequal prediction box technology,combined with multivariate standard normal distribution random numbers and inverse transform sampling.This method first establishes a prediction error power prediction box for the probability distribution of historical wind power data,and uses nonparametric kernel density estimation to fit the error data in the prediction box.Secondly,the multivariate standard normal distribution random numbers are used to inversely transform the error data in the prediction box,and then superimposed on the point prediction to generate a wind output scene.In terms of scene reduction,by calculating the distance between two scenes,selecting a certain tolerance to classify the original scene,selecting a scene from each class with the shortest distance to other scenes to represent the class,this representative sceneCalled a typical scene,its probability is the ratio of the number of scenes of this type to the number of original scenes.By reducing a large number of wind power scenarios,the purpose of streamlining data and reducing the amount of calculation is achieved.Finally,the typical scenarios and point predictions of wind power are applied to the optimization calculation of the day-to-day scheduling model of the power system with wind power,respectively,and the application value of the proposed scenario generation and reduction algorithm in actual engineering is verified.