The calculation of molecular weight distribution (MWD) in conjunction with a computational fluid dynamics (CFD) reactor model is computationally expensive as both the CFD and the MWD require large-scale computations. Operational optimization of polymerization reactors using high-fidelity models like CFD and MWD becomes even more challenging. In this work, a novel method of MWD simulation with a CFD model is first proposed. To overcome the computational expense of the CFD simulations during optimization, an iterative surrogate model method is proposed. Instead of building an accurate and global model with a large set of simulation data, this method uses a limited set of simulation data and updates the optimization iteratively. Optimality of the proposed method is analyzed. A low-density polyethylene production process in an autoclave reactor is used for demonstration. The results show that the proposed iterative method can greatly reduce the computation time and effectively obtain the optimal result.