A comparative study of certain recursive parameters estimation algorithms for multivariable systems
Résumé
A comparative study of parameters estimation algorithms interns of accuracy, model structure complexity, computational time requirements is presented in this paper. The examples include the cases of linear multi-input, multi-output (MIMO) models which cab be decomposed into multi-input; single-output (MISO) representation and interconnected MIMO models. Five MIMO algorithms are considered; the two-stage DU algorithm, the GDU algorithm algorithm, a Prediction Error Model algorithm, the MCA and the MCMV procedures. The last procedure identifies in a recursive way the parameters of a multivariable interconnected system contaminated by colored noise. The comparative assessment is helpful in choosing an appropriate method in a practical case. The advantage of the methods (MCA, MCMV) developed by the authors are clearly given: short horizon of identification, direct treatment of non null operating points, of data sets with missing values, solutions to some non-linear cases when the structure of the process is known.