Research On Fault Classification Method Of Power Quality Analysis

This thesis is mainly concerned with problems of classification of voltage sag, i.e. identification of voltage sag interference source.Power quality is becoming more concerned in recent years, and voltage sag is a common disturbance in electric power distribution system operation. Nowadays voltage sags are probably the most important power quality problem facing many industrial and commercial customers. Voltage sag is a reduction between 10% and 90% in rms voltage, with duration between 0.5 cycle and 1 minute. Voltage sags due to line faults, transformer energizing and induction motor starting are main interference sources. Many methods have been brought forward to resolving this problem in the past., but few of them considering the source classification of voltage sag.For classification of voltage sag, a interference source identification method is brought forward by using wavelet transform and neural network. Characteristics of voltage sag is correspond to certain interference source, so interference source can be identified bysag waveform. According to characteristics of the power system signals, db6 wavelet is used. Using singularity theory of wavelet transform, voltage sag characteristics are extracted easily and accurately by db6. Statistical Learning Theory is a small-sample statistics by Vapnik etal., which concerns mainly the statistic principles when samples are limited, especially the properties of learning procedure in such cases. SLT provides us a novel powerful learning method called Support Vector Machine, which can solve small-sample learning problems better. Support vector machine is based on empirical risk minimization principle, comparing with traditional neural network, support vector machine has not only simpler structure, but also better performances, especially better generalization ability. According to characteristics of the voltage sag, using as the neural network inputs, the voltage sag can be classified by using multi-category classification method of support vector machine combining with decision tree in this thesis.