Filtering identification for multivariate hammerstein systems with coloured noise using measurement data

In this paper, based on the measurement data, the identification of the multivariate Hammerstein controlled autoregressive moving average system is investigated. To facilitate the parameter identification, the considered system is transferred to a regression identification model in which the bilinear parameter and linear parameter are included in the identification model. To solve the bilinear parameter estimation problem, with the help of the hierarchical identification principle, two new identification models are constructed in which the each model is linear to parameter vector. For each identification model, a novel filtering identification algorithm is put forward to interactively estimate the parameters of the each model based on hierarchical identification principle. Filtering technique is used to improve the estimation accuracy of the presented algorithm, and the hierarchical identification idea is exploited to decrease the calculation burden of the proposed method. The conditions of convergence are introduced by using the martingale convergence theorem. Contrast examples indicate that the proposed method has a better identification performance than several existing estimation approaches.