Estimação de estado baseada nos algoritmos LMS em espaço de estado
Keywords: Adaptive filter, SSLMS algorithm, state estimation, state space, synchronous motor
AbstractBased on the state space model, algorithms of type state space LMS (SSLMS - State Space Least Mean Square) allow the generation of an estimated state vector, being a possible solution to the state estimation problem. In terms of tracking ability, the SSLMS is superior to the standard LMS (Least Mean Square), which is limited due to the assumptions of the linear regression model. By overcoming this limitation, SSLMS shows a significant improvement in tracking performance compared to the standard LMS and its known variants. Considering this principle, we propose in this paper the development of the State Space Zero-Attracting LMS algorithm (SSZA-LMS) built around the state space structure. The proposed algorithm is used to estimate the states of a synchronous motor with a nonlinear model. Numerical experiments show that the proposed algorithm presents superior performance when compared to the SSLMS algorithm.