An IV-based method for non-linear continuous-time Hammerstein model identification. Application to rainfall-flow modelling
Résumé
Refined Instrumental variable proved to be a flexible, statistically optimal identification method for discrete as well as continuous-time linear models. This paper presents the extension of these methods to direct non-linear continuous-time Hammerstein model identification from sampled data. The method is not statistically optimal but can provide initial estimates for the initialisation of the optimal prediction error method. After the IV-based method description, some examples are depicted in order to show the performance of the algorithm, the advantages of estimating direct continuous-time models as well as a possible application for rainfall-flow modelling.