Research On The Detection And Recognition Of Power Quality Disturbance Signals

With the rapid development of science and technology,more and more electronic equipment and precision instruments applied to life.It's not only having a higher demand for the power quality,but also bring a very serious adverse effects for the power quality.And the power quality directly affects the economic development of social and the quality of life.Therefore,the identification and governance of power quality is imperative,and receiving more and more attention from domestic and foreign scholars.This dissertation proposed a detection method that uses wavelet transform combined multi-resolution S transform and an identification method named rank wavelet support vector machine(Rank-WSVM)which uses the complex Gaussian wavelet kernel function,which lays the foundation for the governance of power quality.Aiming at long classification time of power quality disturbance signals,a new algorithm based on wavelet transform and multi-resolution S transform is proposed.This algorithm combines the advantages of multi-scale analysis of wavelet transform and the advantages of multi-resolution S-transform which is flexible.Firstly,the original disturbance signal is transformed by wavelet transform to obtain the high and low frequency components.Then the low frequency components were selected to through multi-resolution S transform and extract feature vector.In this way,the characteristics of the original signal are retained,and the size of modulus matrix of this low frequency component after multi-resolution S transform is only a quarter of that of the original signal after direct S transform.This method can not only guarantee the accuracy of classification,but also can greatly improve the efficiency of classification.Considering that the composite power quality disturbance signal is more common in real life.Aiming at Complex disturbance classification,the dissertation proposes the Multi-Rank Wavelet Support Vector Machine algorithm based on complex Gaussian wavelet kernel function.By applying the wavelet technique to Rank-SVM classifier the efficiency is improved.This method is applicable to single disturbance classification and also to the classification of complex signals.At the same time,this method has high recognition accuracy and good anti-noise.Finally,the advantages of the new method can be concluded by simulating this method and applying it to the measured data.The simulation and experimental classification show that the proposed algorithm is effective and feasible.