Research On Signal Processing Technology Of Voltage Sag Rich In Disturbance

Nowadays,the problem of voltage sag has attracted more and more attention.Under the influence of actual loop parameters and operation conditions,the measured voltage sag signals are mostly non-linear and non-stationary,and contain complex noise and many kinds of disturbance signals(harmonics,oscillations,fluctuations,etc.),which makes the detection and identification of voltage sag signals rich in disturbance signals more difficult.The problems of false detection and missed detection of existing methods need to be solved by new methods.For this reason,this theis carries out the research of voltage sag signals riched in disturbance.Firstly,a large number of measured sag signals are analyzed from different perspectives,and a sag analysis method based on space phasor model and improved DBSCAN algorithm is proposed to analyze the characteristics of multi-stage sag signals.By building simulation models,various kinds of sag signals are generated,while superimposing noise and various disturbance signals.Then,the signal source of this theis is formed.Secondly,aiming at the problem that the measured signals may be mixed with harmonics and oscillations,this thesis introduces the Variational Model Decompositon(VMD)algorithm and improves it to separate the sag signal from harmonics and oscillations.Then,the sag amplitude is determined by combining Hilbert transform,and the sag starting and endding time is located by morphological edge detection.In view of the situation that the measured signals may be mixed with fluctuating signals,the traditional extended Prony algorithm is improved,and an adaptive extended Prony algorithm is proposed to fit the fluctuation component in the source signal.Then the adaptive filter is used to remove the fluctuation component and obtain the clean sag signal.Then,aiming at the ten kinds of sag signals which are focused on Chaper 2,a composite sag classification and recognition model based on generalized S transform and multi-level Support Vector Machine(SVM)is built.This model establishes the sag recognition characteristic index system,covering time domain characteristics and waveform characteristics,which are extract by the generalized S transform.Based on this,this thesis constructs a multi-level SVM binary tree to realize automatic recognition of sag types.Finally,this thesis designs a set of signal processing system with high perturbation,which includes two parts: sag signal detection subsystem and sag type identification subsystem.The subsystem of sag detection effectively strips and recognizes noise,disturbance signal and voltage sag signal through a series of signal processing methods.The subsystem of sag identification realizes automatic classification and identification of sag on the premise of effectively stripping noise and disturbance signal.At last the feasibility of the system is verified by simulation signals.