Efficient Algorithms for Time-Domain Dynamic Parameter Estimation

There are several methods for estimating the parameters of a signal in the time domain. Some linear methods are based on polynomials, while others rely on singular value decomposition. These techniques are usually applied to analyze signals offline, that is, as a postanalysis of an event occurring in a physical system. To perform online signal analysis, it is necessary to reduce the number of samples to process by using a sliding window that contains enough information to determine the parameters of the signal of interest. Multiple nonlinear methods have also been developed for online parameter estimation, each defining a data window within its own implementation. This chapter presents state-of-the-art methods for parameter estimation employing the sliding window technique to accelerate the convergence of the linear algorithms. To demonstrate the implementation of the sliding window technique, a simple moving average low-pass filter is employed. To evaluate their performance, two cases are considered. First, an electric power system with a programmable source is used to simulate voltage magnitude and frequency change, and the resulting voltage signal is processed using the proposed algorithms to estimate its magnitude and frequency. Then, one of the proposed methods is successfully incorporated into a control system known as Extremum Seeking Control, an adaptive control technique used to determine optimal inputs or tune systems parameters. In this work, the terms “method” and “algorithm” are used interchangeably to refer to the proposed signal processing techniques.