Research Of Harmonic Detection And Denosing In Electirc Power System

The signal of a power system should contain only sinusoidal signals with a frequency of 50 Hz under the ideal condition.But the reality is often not satisfactory,because there are a lot of external interference,and the equipment has the existence of measurement error,a lot of circumstances lead to the power system signal has the fundamental wave component of the power frequency and some harmonics,even including the noise component.These components pose a great threat to the security and stability of the power system and seriously affect the accuracy of signal processing in the power system.The most serious case is to reduce the service life of power system equipment.In view of the existence of these potential safety factors in the production process can accurately detect the power system harmonic is very important.Firstly,at the beginning of the subject,the original method of harmonic detection in power system,namely Fourier transform,is studied,and how to use Fourier transform to detect harmonic in power system is analyzed.Then,the shortcomings of discrete Fourier transform(DFT),such as asynchronous sampling,spectrum leakage and fence effect,are proposed.Secondly,linear neural network is introduced into harmonic detection of power system because of its simple structure and small calculation.There are two kinds of training algorithms for linear neural networks: least squares(recursive-least-square)and LMS(least-mean-square).Next,the paper puts forward an idea that using wavelet neurons to replace traditional neurons,using wavelet multi-resolution analysis of the power system harmonic signal decomposition,because neural network has arbitrary function approximation ability,so the wavelet transform and neural network,a wavelet neural network to detect harmonic in power systems.Simulation results show that the proposed method has better detection performance.Finally,because there are a lot of in the power system harmonic noise problem,but the traditional threshold functions there are a lot of failed to solve the signal denoising,based on this new threshold function is proposed in this paper,improved the traditional hard threshold function and the defect of soft threshold function,and studies the selection of related parameters.Using MATLAB to experiment,the simulation results of the new threshold function proposed in this paper are compared with the traditional threshold function for signal denoising.Experiments show that the denoising result of the new threshold function is superior.The basis of judgment is Signal-to-noise ratio and mean square error of the signal.