Research On Overvoltage Simulation And Adaptive Identification Of Wind Farm Collection System

As China's scale of wind farm construction continues to expand,the stable operation of wind farm collector systems is increasingly affected by overvoltage.Overvoltage generation can lead to insulation breakdown of electrical equipment and even large-scale power outages.Aiming at the different mechanisms and characteristics of overvoltage generation,a fast and accurate overvoltage identification method is studied,which is of great significance for the rapid maintenance of the collector system,the improvement of insulation and the stable operation of the wind farm.In this paper,by studying the modeling methods of each module of wind farm collector system,combined with the equipment parameters and actual wiring conditions of Hengshan Wind Farm,the overvoltage simulation model of 35 kV collector system is built in ATP-EMTP software,which is frequently generated by the station.Various over-voltage signals are simulated,and the characteristics of various over-voltage signals are analyzed.The effective fault period is intercepted in the time domain as the sample signal to be studied.Aiming at the non-stationary and non-linear characteristics of over-voltage signals,an over-voltage signal decomposition method based on adaptive variational mode decomposition is proposed.A series of eigenmode components are decomposed and orthogonality evaluation indicators are established.Describe the signal decomposition accuracy;considering the difference of the number of modal components after adaptive decomposition of each signal,a modal adaptive unified selection scheme is designed.The time series of each selected mode is divided into several sub-matrices along the time domain to extract time-frequency segmental energy entropy features in each mode,and combined into feature vectors for signal recognition.The support vector machine is applied to the over-voltage signal recognition.The radial basis kernel function is selected by LibSVM,the parameters are optimized by the cross-validation method,and the multi-class support vector machine model is created to realize the one-time identification of various over-voltage signals.Through simulation analysis,it is found that compared with the empirical mode decomposition method,the adaptive variational mode decomposition method has high decomposition precision for over-voltage signals,can accurately extract the characteristics of over-voltage signals,and can unify the eigenvectors of various over-voltage signals.Dimensions.Compared with the traditional energy entropy characteristics,the difference between the time-frequency segmental energy entropy characteristics of various over-voltage signals is more obvious,which is beneficial to classifier identification.The SVM multi-classifier application built by LibSVM can effectively improve the recognition accuracy in various types of over-voltage identification,and the calculation rate is high.The method proposed in this paper satisfies the actual needs of automatic identification of over-voltage in power systems,and provides technical support for future engineering applications.