| For reliable power system operation, the ability of a system to maintain synchronism after a disturbance is greatly important. This small-signal stability is governed by inter-area electromechanical modes, which dictate how generators in disparate parts of the system interact electrically and mechanically. Several measurement-based mode-monitoring methods have been developed in recent years. Unfortunately, these algorithms suffer severe bias when rogue system inputs known as forced oscillations are present in the data. These forced oscillations arise at generation sites and often impact measurements from throughout a system.;To address the impact of forced oscillations on mode estimates, an adapted system model that incorporates forced oscillations was developed. Based on this model, the Two-Stage Least Squares and Overdetermined Modified Yule-Walker algorithms were redeveloped, resulting in the LS-ARMA+S and YW-ARMA+S algorithms. These new algorithms are capable of simultaneously estimating a power system's modes and the amplitudes and phases of forced oscillations. To properly incorporate forced oscillations into the system model, algorithms to detect the forced oscillations and estimate their frequencies, starting points, and ending points were needed. For this purpose, a classical periodogram-based approach for the detection of sinusoids in white Gaussian noise was adapted to operate on power system data. The resulting algorithm is capable of detecting sinusoids embedded in colored noise, is optimized for use in the online environment, and provides frequency and amplitude estimates for each forced oscillation. To estimate the starting and ending points of the forced oscillations, a novel time-localization approach based on the sample cross-correlation between the measured data and a complex sinusoid was developed.;To test the proposed methods, they were applied to simulated and measured power system data. Results indicate that forced oscillations can be reliably detected and incorporated into the system model. As a result, the redeveloped mode monitoring algorithms exhibit improved performance compared to their original counterparts. The methods are shown to be a viable approach to estimating a power system's electromechanical modes in the presence of forced oscillations. |
