Power Quality Analysis Based On Wavelet And HHT

With extensive use the non-linear power elements in power system, the power quality causes more and more attention. The harmonic identification, classification different type of power quality disturbances and the de-noising are studied in the thesis. The main contains are as fellows:1. A new theory Hilbert Huang Transform (HHT) is put forward to analysis the harmonics of power system due to the serious harmonic problems nowadays. The method how to identify the parameters of harmonic is given. This method can overcome the spectrum leakage of Fast Fourier Transform (FFT) and it can also deal with the shortcoming of choosing the wavelet when analysis the harmonic. Furthermore, it is complete a data-driving method, can decompose the harmonics automatically from the power quality disturbance signals, and then realizes identification harmonics. In order to achieve more accurately harmonic analysis, the new interpolation function-piecewise Cubic Hermite function and different kinds of technique are adopted to alleviate edge effects.2. Classification various power quality disturbances are vital to make effective decision to deal with the corresponding problems. Feature vectors extracting is very essential for classification. This thesis put forward a new way to obtain it. The feature vectors can embody the characters of power quality disturbance in different frequency bands, and its dimension is fairly low which is more suitable for classifier to classify. Least Square Support vector machine is introduced to identify different kinds of disturbances. It can overcome the shortcoming of neutral network, which is based on empirical risk minimize, need large number of training samples and easily get into local minimum.3. The power quality signals corrupted by large mount of noise can greatly affectthe result of detection in the power system. This thesis improves the algorithms based on Minimum Description Length (MDL) criterion. The second term in MDL function is changed to the signal retained energy ratio, which can improve the ability to de-noise, and then the corresponding parameters in improved MDL are given. Moreover, the wavelet basis function selection and the decomposition levels are discussed in detail. The method in this paper is superior to the wavelet and MDL based de-noising method. The result is satisfaction.Large mount of simulations indicate that the wavelet and HHT have the probability and the validity to analysis the power quality problems.