The Detection And Classification Of Transient Power Quality Disturbances Based On Complex Impedance And Fuzzy SVM

Electric power is gradually becoming the basic energy with the rapid development of national industrialization level and science, and the requirement of power quality is also gradually improving in different sectors. In addition, due to the widely application of a variety of nonlinear load, impact load and fluctuation load, the power quality disturbances occur frequently with the growing loss of the social economy. As the embodiment of the national industrialization level, it is a popular topic that realizing the elimination of power quality disturbances(PQDs), improving the power quality and building high quality electric power network. But, it is still difficulty governing the disturbances for the wide varieties of PQDs and complicated actual situation.The accurate classification of PQDs is the precondition of eliminating PQDs and the primary guarantee of improving power quality. This paper presents the support vector machine and the fuzzy support vector machine based on the complex impedance, and setting the PQDs feature extraction, optimization and classification with effective results. For this paper, the main work is as follows:1. This paper proposes a feature extraction method based on the complex impedance. Firstly, Hilbert transform is used to construct the analytic signals for the voltage and current signal separately, and then their corresponding complex impedance can be calculated, then educe the disturbance characteristics such as:the difference of complex impedance modulus value, the phase jump maximum of complex impedance, the residual value, the parameter of residual value and the duration with the software phase-locked loop to constitute feature vector. Finally, the experimental results show that this method easy to accomplish, extract the disturbance features effectively, it has wide promotional value.2. This paper introduces the Fuzzy Support Vector Machine (F-SVM), aimed at the disadvantages of the SVM. Firstly, we build the membership function based on the Bayesian theory and the concept of sample density. Then, the fuzzy membership of each point would be confirmed, and this method reduces the negative effects of noise in the dataset. Finally, the experimental results present a better accuracy and robustness in complicated environment.3. Firstly, this article makes comparisons among different methods under different noise to find out the advantages and disadvantages of different ways for feature extraction, the methods include complex impedance& SVM, complex impedance& F-SVM and S transformation& SVM. Then we compare the different classification methods under the white noise and it analysis the precision between different classifier under the same method of feature extraction.