A multi-model structure for model predictive control

Model predictive control (MPC) is a wide popular control technique that can be applied starting from several model structures. In this paper black-box models are considered. In particular it is analysed the sets of regressors that it is better to use in order to obtain the best model for multi-step prediction. It is observed that for each prediction a different set of real data output and predicted output are available. Based on this observation a multi-model structure is proposed in order to improve the predictions needed in the computation of the MPC control law. A comparison with a classical one-model structure is discussed. A simulation experiment is presented. © 2004 Elsevier Ltd. All rights reserved.