On-Line Disturbance Signal Detection And Low Frequency Oscillation Modes Estimation Based On WAMS

With the continuous development of power grid, potential problems in securityhave been brought by. Disturbance of power grid and low frequency oscillation surfacefrom time to time, which become the critical factors that limit power transfer andjeopardize the safety and stability of grid operation. Safe and reliable operation ofpower grid faces new challenge, it is very necessary to detect the disturbance of powergrid and estimate low frequency oscillation modes. Thus, based on wide areameasurement system and the index of normalization kurtosis, adopting the probabilitydistribution characteristics of the data from phasor measurement unit, a method fordisturbance signal on-line estimation and a new scheme for low frequency oscillationmodes on-line estimation are presented in this paper.This paper presented the common autoregressive moving average (ARMA) methodthat is suited for analyzing ambient signal, and the Prony method that is suited foranalyzing ringdown signal. The deficiencies of the common scheme, which combinesARMA method and Prony method, are the bad performance of disturbance detectionmethod and the problem of algorithm-switching.Based on the index of normalization kurtosis, the probability distributioncharacteristics of the standardization PMU signal is analyzed in this paper. The testresults of several measured signal of power grid show that when the normalizationkurtosis value of ambient or ringdown signal is almost3, the signal is approximategaussian signal, but the normalization kurtosis value of mixed-signal which contains thetwo kinds of signal is much greater than3, it is super-gaussian signal. Based on theabove research and the technique of sliding window, a new method for disturbancesignal on-line detection is presented, this method calculates the normalization kurtosisvalue of the standardization PMU signal, which is updated real time by the slidingwindow, and compares the result with the disturbance threshold, then determineswhether the ringdown signal exists in the sliding window or not.In order to solve the deficiencies of the common scheme, the high order ARMAmethod, which is suited for analyzing the mixed-signal, and a new scheme for lowfrequency oscillation modes on-line estimation are presented. Based on normalizationkurtosis，it can judge whether ringdown signal exists in the sliding window or not, andcan self-adaptively switch between the common ARMA method and the higher order ARMA method. According to ambient signal or mixed signal existing in the slidingwindow, it can adopt the corresponding method to the signal, so it could ensure thereliability and effectivity of low frequency oscillation modes on-line estimation.Case studies are implemented based on measured signal of power grid, the resultsdemonstrate the validity of the proposed method for disturbance signal on-line detectionand the proposed scheme for low frequency oscillation modes on-line estimation.