This paper proposes an offset-free model predictive control ( MPC) algorithm for constrained nonlinear systems, by utilizing an artificial disturbance to replace the modeling error. The algorithm is based on a new expanded system structure and the dual part control law consisting of the linear stabilizing control and the dynamic predictive control. Compared with previous techniques, there is no need for calculating the steady state input, state targets and using an observer for augmented states. Other characteristics of this algorithm include controllable expanded system model and handling model mismatch. By consideration of the state and input constraints, design of the MPC controller not only achieves the aim of offset free control but also guarantees the constraint satisfaction in the presence of disturbance and model mismatch. The particular characteristics of the pro posed algorithm are illustrated via a simulation using a continuous nonlinear jacketed stirred tank reactor( CSTR) model.