Study On The Theory And Technology Of Real Time Power Quality Analysis

With the growing complication of power grids, increasing usage of power electronic devices and other nonlinear loads, power quality issues get more concerned than ever before. Scholars and engineers have done many researches on it and achieved many feats. Power quality analysis mainly includes three aspects:power data compress, power quality disturbances classification and disturbance parameters estimation. In different usages, the requirement of the three aspects is different too. Many complicated methods have achieved better performances than the methods proposed by IEC/IEEE on accuracy. However, with the rising accuracy, runtime is also sacrificed. Currently, the real time ability is required more and more, and in order to investigate high frequency disturbances, sample rate is increasing too. Therefore studies on high accuracy and effectiveness power quality analysis methods are meaningful and necessary.With the premise of good accuracy, this dissertation takes the consideration of real time ability as well, and proposes effective methods on power data compress, power quality disturbances classification and disturbance parameters estimation. The dissertation is divided in to the following issues:(1) Design an online and real time ability power quality disturbances detection and classification device which is based on DSP-FPGA platform to realize sign sampling and processing. The design provides a real time ability test platform for the proposed methods.(2) Propose a parameterization power data compress by using strong trace filter, Dynamics and lifting wavelet transform. The strong trace filter cannot only give effective parameter estimation, but also recognize the transient disturbances clearly; Dyn can recognize interharmonic under noise; lifting wavelet transform can compress the signal as discrete wavelet transform do with half runtime and memory. By subtracting the fundamental and harmonic components from the signal, the compress ratio of transient and interharmonic disturbances has been increased greatly by using the lifting wavelet transform. The proposed method has better accuracy, reliability and practicability than the related methods. (3)Time-frequency domain analysis is an important feature extraction way of power quality disturbance classification. The dissertation proposed a hybrid method based on S transform and Dynamics, which ensures a high accuracy while maintain a very low runtime. Heisenberg’s uncertainty is a bottleneck of many time-frequency domain analysis methods. The dissertation proposed a frequency varying Gaussian window which decreases the influence of Heisenberg’s uncertainty, and raises the accuracy of feature extraction. Five distinctive features are selected for a deliberated decision tree for the classification. Each branch of the tree only has one feature as judgment, which avoids ambiguous errors. The scalability of decision tree allows users to satisfy their own interest by easy modification on the tree. The compatibility of the proposed method is increased.(4) Disturbance parameter estimation is one of important branches of power quality analysis, and the criterion to evaluate the power quality. Nowadays, many good works have been down on stationary disturbances while the transient disturbances still leave some problems to be solved such as accuracy, influence from frequency deviation, detection delay and so on. The dissertation forces on the most common transient disturbances sag/swell, and proposes a real time sag/swell detection method based on strong trace filter. The proposed method is with low computation burden, immunity from frequency deviation, short delay of detection. The proposed method is not only suitable for sag/swell detection, but also for sag/swell restoration purpose.