Local Identification of LPV Dual-Rate System With Random Measurement Delays

Abstract

This paper considers the problem of linear parameter-varying (LPV) dual-rate system identification with slow-rate output data corrupted by random measurement delays. The local approach is used, and the LPV model is built through convex combination of several local linear models with time-varying weights. The dual-rate sampled data often exist in industry, and the measured output data are inevitably corrupted by random measurement delays. In order to handle random measurement delays and dual-rate sampled data in LPV system identification, the statistical description of the identification problem is established based on the presented output-interpolated LPV dual-rate model, and then, the proposed algorithm to estimate all the unknown model parameters, real output data, and measurement delays are derived in the generalized expectation-maximization algorithm scheme. One simulation example and a practical chemical process are used to verify the efficiency of the proposed method.