Nonlinear Analysis Methods Of Synchrophasor Measurements For Fault Diagnosis In Power Systems

The continuous supply of electrical energy is the basic guarantee for the rapid development of modern society and economy.According to the latest renewables global status report,total global renewable power capacity was up to nearly 2,017 GW at the end of 2016.With the continuous introduction of distributed renewable energy generation and the expanding scale of the power system,the structure and operation mode of the power system become more complex.It poses a higher challenge to the stable operation of the system.In order to solve the above problems,Phasor Measurement Unit(PMU)is used to collect real-time information of each area of the entire network.It provides data sources for failure analysis.This paper studies timely and effective nonlinear diagnosis methods to achieve rapid fault detection and accurate fault location.So,it has important research value and practical significance for stable operation of the power system.Based on PMUs data,this paper introduces Kernel Principal Component Analysis(KPCA),Hessian Local Linear Embedding(HLLE)and Nystr?m approximation theory into power system fault diagnosis.Resorting to these ways,some new nonlinear fault diagnosis algorithms proposed can improve accuracy of fault location,reduce the computational complexity and process time-varying characteristics of data.At the same time,MATLAB is used to complete the corresponding simulation experiments that obtain good results.The main work and innovative points are listed as follows:(1)The development status of the power system fault diagnosis methods is reviewed.The basic principles of PMU and its application in power systems are analyzed.This paper studies the existing fault diagnosis methods in power systems,and analyzes the limitations of algorithms to provide technical support for theoretical exploration and performance analysis of new algorithms.(2)Most of data-driven methods for fault diagnosis in power systems are linear transformation methods.They are difficult to effectively extract nonlinear characteristics of PMU data.So,accuracy of fault detection and fault location will be affected.The KPCA method is introduced into power systems for fault location.The partial derivative of the polynomial kernel function is deduced by means of the scale factor.Then,the contribution of each variable to the statistic is derived to determine whether a bus is the fault component and evaluate its propagation across the system.Based on KPCA of PMU data,a nonlinear method is proposed for fault location in complex power systems.Compared to the Principal Component Analysis(PCA)-based method,the novel version can provide a lower false alarm rate and a better positioning effect.(3)The KPCA method requires complicated nonlinear mapping,which increases the computational complexity of the algorithm.Therefore,the application of HLLE theory which can effectively compress massive data provided by PMU is explored for fault detection in power systems.The local linear regression is used to find the projection that best approximates the implicit mapping from high-dimensional samples to the embedding.It can reduce the computational complexity of processing new data.Based on the above studies,a power system fault detection method based on HLLE is proposed.Compared to KPCA-based power system fault detection method,the proposed algorithm reduces the detection time while provides almost the same detection performace.(4)PMU data has a time-varying characteristic.As the KPCA model and its threshold is not adaptive,new stable operating conditions may be classified as faults.This paper uses the moving window to solve above problem.For the high computational burden involved in the kernel matrices construction and the related eigenvalues decomposition of KPCA algorithm,the Nystr?m approximation method is used to realize the approximate reconstruction of the kernel matrix,principal eigenvalues and eigenvectors.A new method of fault diagnosis in power systems is proposed.Compared to Moving Window KPCA(MWKPCA)-based power system fault detection method,the novel version reduces the computational complexity by 37% while provides almost the same detection performation,and can provide the precise identification of fault location.This paper explores KPCA method,HLLE method and Nystr?m approximation to cope with fault diagnosis problem in power systems.It will provide technical support for stable and safe operation and real-time monitoring and protection of the power system.