Research Of Harmonic Detection And Denosing In Electric Power System

In an ideal situation, power system signal only contained sine signal of50Hz. However, in theactualengineering application, therewere outside interference and error of measurement of the equipment.Sopower system harmonicsignals wereusually consist of fundamental components, every har-monic components and noise components. On the one hand, the existense of every harmonic compo-nent and noise components can have a serious effect on the processing accuracy of power system signal. On the other hand, the pollution of harmonic and noise caused great potential threat to the safeand stable operation of power system. At the same time, they also shorten life of equipments in thepower system. So it was very necessary and meaningful to give detection and denoising processing to the harmonic of power system.Firstly, this thesis studied the traditional Fourier transform. It can detect power system harmoni-c signals.And then analysized the fundamentals of the Fourier transform for detecting harmonic sig-nals. According to defects of the Fourier transform (i.e. asynchronous sample, spectrum leakage and fence effect), short time Fourier transform was studied. And the power system harmonic detectionmethod based on Nuttall’s window interpolation algorithm was put forward. The simulation resultsshowed that window function and interpolation algorithm can effectively solve the defects of the traditional Fourier transform and also to improve the measuring accuracy.Secondly, the basic principle of wavelet transform was studied. According to the low accuracyproblems of harmonic detection based on Fourier transform, power system harmonic detection met-hod based on wavelet neural network was put forward. The related parameter selection in the wavelet neural network was studied. And the improved steepest descent method was applied to train the n-etwork. The simulation results showed that the method can be more accurate and fast in power system harmonic detection.Then, because of the harmonic noise problem and the traditional deficiency of threshold function in power system, harmonic denoising method based on asymptotic half soft threshold function was proposed. The basic principle of the wavelet threshold denoising was studied. The simulation results showed that harmonic denoising method based on asymptotic half soft threshold function can o-btain higher signal-to-noise ratio and smaller root mean square error.