The Research On Voltage Fluctuations And Flicker Detection Based On Wavelet Transform

Two important parts are included in voltage fluctuation and flicker detection, which are extracting voltage fluctuation signal envelope and effectively analyzing the extraction voltage fluctuation signals, and confirming short-term flicker severity. The extraction of the signal in voltage fluctuations is the premise to analyzing the voltage fluctuation, and the analysis of voltage fluctuation signal is the key to examine the voltage fluctuation and flicker.The flicker signal extracted is analyzed in this thesis.Discrete Fourier transform and wavelet transform are two methods usually used in the analysis of flicker signal. DFT has the shortcoming of spectral leakage and the fence effect, which will lead to the calculation accuracy of Pst is not high, while the wavelet transform with the features of time - frequency localization, which overcome the drawbacks of the DFT, is very suitable in the analysis of mutant signal, and it can detect the start time and end time of flicker signal, so wavelet transform is used to analyze the voltage fluctuation and flicker in the thesis.Wavelet transform is based on the theory of multiresolution analysis which use Mallat algorithm. Mallat algorithm exist problem of frequency folded and energy leakage.To resolve the problem of frequency folded, DFT and IDFT algorithm are combined to form DMallat algorithm in this thesis, with the comparison test to the calculation of short-term flicker severity P_st using DFT and DMallat, DMallat's effectiveness and stability is shown, and calculation accuracy of short-term flicker severity P_st is significantly improved.Selecting the wavelet function properly or selecting the sampling frequency reasonable can reduce the energy leakage. DbN wavelet is selected in the thesis.Energy leakage can be reduced by increasing N value but with the computation increased at the same time, in order to reduce the computational volume to a single frequency flicker signal, down-sampling points method is used in the thesis. To multi-frequency flicker signal, through many simulation test frequency range of options is given to reduce the energy leakage, that baseband signal is near to the minimum of frequency band approximate part of wavelet decomposition as far as possible and harmonic component part is located in the center of the details.In the description of the relationship between unit value of instantaneous flicker frequency and the corresponding frequency sine voltage fluctuations value, the BP network of an excellent ability of nonlinear function approximation and interpolation properties is employed on setting model of frequency and sin voltage fluctuation value at unit short term flicker to improve fitting accuracy.