Blind Source Separation Based Modal Analysis Method For Low Frequency Oscillation

Nowadays,on-line low frequency oscillation(LFO)monitoring and analysis receive more and more attentions,owing to an increasing interest in the prevention of grid breakouts and widespread blackouts,which can result from uncontrolled power swings across large geographical areas.There are two issues of on-line LFO monitoring and analysis: determining from the measured power system signal whether/when the LFO occurs;then qualifying the underlying threat of LFO by effectively assessing the modal parameters(oscillation frequency and damping coefficient)of different dominant modes.To fulfill on-line monitoring and analysis task of LFO,this paper proposes a blind source separation(BSS)based modal analysis method.This is the first time when BSS technique is applied as a modal decomposition tool for LFO modal analysis.At first,this paper introduces the BSS technique and suggests that it can accomplish modal decomposition of LFO signal.Afterwards,a BSS based modal analysis method for one-channel LFO measured signal is proposed.The procedures of the proposed method are as below: Firstly,pre-process the one-channel measured signal,including applying mathematical morphological gradient(MMG)operator on the measured signal to detect the onset instant of LFO and establishing multi-channel measured signal matrix by employing Takens' embedding theory;Secondly,utilize BSS technique,SOBI algorithm,to decompose multi-channel measured signal matrix into different mode signals;Thirdly,adopt modal parameter estimation tool,HT technique,to assess the oscillation frequency and damping coefficient of different modes.To solve the problems that the numbers of LFO modes(i.e.model order)is unknown before BSS and dominant modes should be identified from decomposition results within noisy circumstance,this paper proposes a method,SOBI-HT based Dominant Modes Identification(SHDMI)method,which combines an iteration procedure with an identification procedure,to determine the optimal value of model order,and then identifies the dominant LFO modes.The performance of SHDMI method is evaluated on synthetic signals,simulation signals and a real-time measurement data,compared with Prony and HHT.Analysis results verified that SHDMI method can act as a novel and reliable tool,which is capable of on-line monitoring and automatically analyzing one-channel measured system response signal,to accurately assess the stability stage of power system.