A new approach for robust model predictive control with an application to an air-conditioning system

A new approach for modeling and controlling a class of hybrid systems described by interacting physical laws, logic rules, and operational requirements is proposed. As recently suggested in the literature, logic rules and operational requirements are integrated into the model as equivalent linear inequality constraints involving both continuous and binary variables. The distinguishing feature of the approach proposed is the use of orthonormal series for modeling the process dynamics. The model, which is referred as MLDOSF-mixed logical dynamical, based on orthonormal series function - can be easily estimated with only approximate knowledge of the system dynamics. The orthonormal basis is able to represent a stable system with fewer parameters than in a finite impulse response model. Uncertainties in the process dynamics are incorporated into the output equation only, reducing the model politopic space. A robust predictive control sequence is obtained by optimizing a performance index in a min-max sense, subject to the model constraints. An application to an air-conditioning control system illustrates the main characteristics of the MLDOSF framework proposed.