On the physical meaning of minimization criterion for Model Predictive Control

The present work is aimed to investigate the minimization criterion used in Model Predictive Control (MPC). The paper shows that minimization of currently used objective function in MPC does not minimize the consumed energy in the case of true bond graph models of the systems. Thus, it is proposed to use another criterion which removes this drawback. A consequence of this change is that the algorithm does not cancel the steady-state error of the system. Therefore a cancelation of the offset error by tuning the weighting coefficients is proposed.