Classification of power quality disturbances based on principal component analysis and support vector machine with optimal kernel-parameter
It proposes a method to identify Power Quality Disturbances (PQD) with noise based on wavelet energy difference and Support Vector Machine (SVM) of optimizing the nuclear parameters. PQD signals are decomposed into 10 layers by db4-wavelet with multi-resolution. Energy Differences (ED) of every level between PQD signal and standard signal is extracted as eigenvectors. Principal Component Analysis (PCA) is adopted to reduce the dimensions of eigenvectors and find out the main structure of the matrix, which forms new feature vectors. Then these new feature vectors are divided into two groups, namely training set and testing set. The method of cross-validation is used for the training set to select the optimal parameters adaptively and construct the training model. Finally, the testing set is substituted into the training model for testing. The results show that the method has high resolution, strong resistance to noise, fast classification speed, and is suitable for the classification of PQD.
Author's Name: Liu, G., Li, F., Zheng, S., You, Y.
Volume: Volume 11
Issues: Issue 2
Keywords: Classification, PCA, Power quality, SVM, Wavelet energy difference