Using Adaptive Step-Size Least Mean Squares (ASLMS) for Estimating Low-Frequency Electromechanical Modes in Power Systems

Information about the stability of heavily interconnected power systems is found in low-frequency electromechanical modes. This research complements model-based and measurement-based approaches to mode estimation typically requiring much computation and ringdown from a disturbance, respectively. This paper discusses an adaptive step-size least mean square algorithm (ASLMS) for estimating the electromechanical modes near real time in which the step size (mu) is adaptively controlled to improve the convergence time of the estimates compared to the LMS. A simple gradient vector (gamma) and a small positive constant (rho) are introduced for controlled adaptation of mu. The improvement in convergence performance of the frequency and damping estimates when compared to the LMS estimates is shown by applying the ASLMS to simulated data containing a stationary mode from a 19-machine test model and ambient power system data, while using an initial weight vector from previous ASLMS, LMS and block processing algorithm estimates