Wind Farms Output Models Considering Dependence And Its Applications In Power System

In the rich region of wind resource, the distance in adjacent wind farms is generally not more than 10 kilometers. Wind farms in the field have the same wind speed, basically. Then, the output of adjacent wind power will have a certain correlation. For the strong randomness of wind power, the calculation of power flow may result large errors if the correlation was neglected. As the exists of correlation, wind farms' output may affect the safety and stablity of power grid operation. For the probabilistic power flow problem of large wind farm access system, a new probabilistic power flow calculation method, which was based on the short term dependence model of single wind farm, the correlation model of multi-wind farm output and the correlation of input variables, was presented.Considering assumption that the traditional wind speed parameter model had the characteristics of the error and the self-correlation of wind speed, a new method based on non-parametric kernel density estimation and Markov theory was proposed. Wind speed distribution obtained by non-parametric kernel density estimation could reveal the potential statistical information, and there was no hypothesis error. Moreover, Markov theory could characterize the time series of variables. The numerical results showed that the model could describe the distribution law of historical wind speed more accurately, and the short term dependent property of the wind speed also be maintained.In view of the error between empirical Copula and actual data model, a method of constructing the joint distribution of multi-wind farm output based on Copula kernel function was proposed. Using kernel density estimation and Copula function as the theoretical basis, a continuous Copula kernel function was derived, and the optimal Copula function for the output distribution of the wind farm could be accurately described by Copula kernel function instead of Copula. The Copula kernel function derived from this method based on the sample data, which could effectively reduce the error between model and actual data, effectively. What's more, the Copula had some better mathematical properties. The results of an example showed the effectiveness of the proposed method in this paper.In view of the existing probabilistic load flow calculation method, a new method based on Copula function was proposed to deal with the nonlinear dependent input variables. The method was generated to generate some random samples by Copula function, and then combined with the edge of wind farms' output to get power samples.In order to further explain that the kernel Copula was more effective than empirical Copula, the Copula function generated by two methods to satisfy the relevant conditions of wind farms' output sampl. The power flow results in IEEE30 node test system showed that the Copula based on power flow results were more close to real power flow. The model established in this paper provided a new way to analyze the influence of wind farms output.