Abstract: Model uncertainty can deteriorate the steady state objective in a two-layer model predictive control system. Based on the economic optimization model with a point model, we propose an iterative compensation method to correct the equality constraints of the point model, considering the change of the inputs' increment and the errors of the outputs' original values, and to update the optimal set-points. The calculation of optimal set-points becomes an asymptotic, dynamic transient process. The simulations show that the error of the steady-state objective is largely diminished and the influences of model uncertainty on steady-state optimization are improved effectively.