Abstract:
Information about the location and strength of low frequency electromechanical modes in power systems reflects the stability of the system. Highly recommended and used techniques like Prony analysis and eigenanalysis require ring down from a disturbance and tedious matrix calculations, respectively, for mode estimation. This work proposes the use of the Least Mean Squares (LMS) adaptive filtering algorithms and its combination with other algorithms for estimating and tracking the modes with respect to time. The mode of interest in this work was the 0.26 Hz mode. An Adaptive Step Size Least Mean Squares (ASLMS) algorithm was introduced in this work to reduce variability in mode estimation for non-stationary environments. The ASLMS algorithms achieved quicker convergence than LMS algorithms. A combination of the ASLMS and the LMS algorithm called the Error Tracking (ET) algorithm was tested, based on the running error in the estimate, to reduce variability while also maintaining reasonable convergence time. The ET algorithm achieved high accuracy, less variable performance and quicker convergence of estimates compared to all the other algorithms. The ET algorithm tracked the 0.26 Hz mode in both the simulated data and the real time data with the least amount of error.
