Power quality disturbances classification Based on

this paper used the multi-class classification for support vector machine and combined with the good amplitude-frequency characteristic of fourier transform,the good time-frequency characteristics of wavelet transform and the excellent statistical learning ability of support vector machine to make the classification and recognition to the disturbances of power quality. Mathematical modeling for the 8 kinds of common power quality disturbances, namely voltage swell, voltage sag, voltage interruption, harmonic, voltage fluctuation, transient oscillation , transient pulse and frequency deviation, and then use fourier transform and wavelet transform to extract the characteristics of the waveform of the generated samples, and input the characteristic value to the osu_svm and do the quality disturbances Multi-class Classification. The example shows that this method has a high recognition accuracy, a few training samples and training time is short, a good real-time performance, and is not sensitive to noise, etc. It is an effective method for Power quality disturbances classification. Key words-power quality; disturbances classification; support vector machine; multi-class classification I. INTRODUCTION The problem caused by the power quality disturbances has been increasing concern, in order to reduce the impact of the disturbances, we must to detect and identify the problem firstly and provide the value information for the problem's diagnosis and solution. And this requires to classify the various disturbances correctly, but as the power quality monitored date is very large, and it is a time-consuming and laborious work for the skillful worker to classify them. Therefore to achieve a automatic classification to the various disturbances accurately and fast has become a hot topics of current research in the field ofpower quality analysis'!', At present, most power quality recognition methods all have used artificial intelligence approach to power quality disturbances to achieve automatic recognition, such as wavelet transformation, the fuzzy expert system, the union wavelet transformation and FFT, the multi-dimensional minute,and artificial neural network.The neural network has the characteristics with simple structure and strong ability to solve the problem, and it is an important recognition method, but it has some shortcomings such as the local optimal problem of algorithm, the training time is long ,and easy to over-fit and ect. The fuzzy technology forms the judgment and has a higher efficiency of identifing through the knowledge rule of the simple "IF-THEN" form, but as a result of many power quality disturbances, for example the overtone, the vibration, the voltage undulation and so on, and it is very difficult to establish "IF-THEN" such obvious knowledge rule,