Multi‐Innovation Gradient Identification Methods for Bilinear Output‐Error Systems
Meihang Li, Ximei Liu, Yamin Fan
Abstract
ABSTRACT This article addresses the parameter estimation problems of bilinear output‐error systems, and the auxiliary model identification idea and the particle filtering technique are adopted to overcome the identification obstacle resulting from the unknown true outputs. Then a particle filtering‐based forgetting factor stochastic gradient algorithm is proposed for the identification of bilinear output‐error systems. To enhance the convergence rate and accuracy of parameter estimation, we expand the scalar innovation to an innovation vector and develop a particle filtering‐based multi‐innovation forgetting factor stochastic gradient algorithm. Finally, a numerical example and a practical continuous stirred tank reactor process are provided to show that the discussed methods are work well. The results indicate that the proposed algorithms are effective for identifying the bilinear output‐error systems and can generate more accurate parameter estimates than the auxiliary model‐based forgetting factor stochastic gradient algorithm.