Power Quality Analysis And Load Modeling Based On Support Vector Machines

Statistics is to infer the law of nature according to observation data. Statistical learning theory is a newly developed theory for studying the statistical estimation and prediction problem based on small number of samples. It studies the nature of machine learning.As the latest development of statistical learning theory and the embodiment of structural minimization criterion, Support Vector Machines (SVM) provide efficient and powerful algorithms that are capable of dealing with high dimensional input features and with theoretical bounds on the generalization error and sparseness of the solution provided by statistical learning theory. SVM has few free parameters requiring tuning, is simple to implement, and are trained through optimization of a convex quadratic cost function, which ensures the global optimization of the SVM solution. Furthermore, SVM-based solutions are sparse in the training date and are defined only by the most"informative"training points. SVM presents a lot of advantages for solving the small samples, nonlinear and high dimensional pattern recognition, as well as other machine-learning problems such as function fitting. The thesis studies the SVM in the applications of power quality analyzing, power quality modeling and load modeling in order to perform some innovation for these hard problems.In detail, the major contributions of the thesis are as following:A novel robust algorithm to interharmonic analysis based on SVM and solved by Iterative Reweighted Least Squares algorithm to overcome its difficulty of exponential computation complexity, is proposed in the paper. There is a good precision for analyzing harmonics and interharmonics without synchronized sampling. By introducing a specific loss function, the method can mitigate the infection of outliers and noises and exhibits robustness characteristics. Its IRWLS–based implementation makes it efficient and suitable for harmonic and interharmonic analysis of electric power system. The case studies showed its high precision and robustness of the SVM spectral analysis algorithm. Based on SVM and S-transform, a novel scheme to detect and classify various types of electric power quality disturbances is presented. The S-transform is an extension of the continuous wavelet transform and short time Fourier transform, it uses an analysis window whose width is decreasing with frequency and then providing a frequency dependent resolution. For its good time-frequency characteristic, it is suitable for feature extraction of power quality disturbance signals. At first the S-transform is applied to obtain useful features of the non-stationary power quality disturbance signals. Then disturbance types are identified through the pattern recognition classifier based on SVM. Numerical results show that the proposed classification method is an effective technique for building up a pattern recognition system for power network disturbance signals. A novel concept of dynamic power quality disturbance classification tree is also put forward to tackle the open problem of multiple disturbance classification. This concept is based on the Mercer Kernel clustering algorithm that dynamically expand the static power quality classification tree.SVM is utilized to establish harmonics load model, static load model and dynamic load model. In harmonics load modeling, SVM is utilized to construct nonlinear mapping of magnitude and phase angle of every order harmonics current to magnitude and phase angle of every order harmonics voltage and harmonics load parameter. In static load modeling, the nonlinear mapping between active power or reactive power and voltage magnitude, frequency and load parameters is established by SVM. In dynamic load model, the nonlinear mapping between dynamic active power or dynamic reactive power and dynamic input vector is established by SVM.The power quality monitor and analysis platform is development and validated by dynamic simulation experiment. The platform can implement comprehensive monitor and analysis for power quality.