Multi-Model Approaches for Bilinear Predictive Control

It is well known that linear controllers can exhibit serious performance limitations when applied to nonlinear systems since nominal linear models used during design cannot represent the nonlinear plant in its whole operating range (Arslan et al., 2004). For this reason, several researches has been proposed new techniques in order to supply a solution for this problem. The main alternative technique, proposed by academy, to resolve the referred problem is known as multi-model approach. The basic idea of multi-model approach consists in decompose the system’s operating range into a number of operating regimes that completely cover the chosen trajectory as showed in (Foss et al., 1995). There are, basically, two approaches for multi-model. The first one consists of to design a set of suitable controllers (one for each operating regime) and to calculate weighting factors to them as showed in (Arslan et al., 2004) and (Cavalcanti et al., 2007a). The global control signal is a weighting sum of the contributions of each controller. The second one consists of to build a global model as a weighting sum of each local model as showed in (Foss et al., 1995) and (Cavalcanti et al., 2007b). In both cases, a way to measure distances between models is defined. Multivariable Model Predictive Control (MMPC) has been presented in this chapter. MPC is the an of the most important control technique used in industry. Multivariable Bilinear Generalized Predictive Control (MBGPC) is formulated and, its alternative solution, Multivariable Bilinear Generalized Predictive Control with Iterative Compensation (MBGPCIC) is presented. This chapter shows either proposed metrics in order to build multi-model based controllers (based in MBGPC and MBGPCIC) and presents simulation results applied in distillation columns.