Abstract: A preference-based non-dominated sorting genetic algorithm (PNSGA) was proposed to solve the multi-objective optimization problems. It is an improved method of NSGA Ⅱ(non-dominated sorting genetic algorithm Ⅱ）. A new preference relationship was defined based on Pareto dominance and goal vectors. The goal vectors were designed according to the decision-maker’s preference. This algorithm combined the preference relationship with the fast non-dominated sorting, an important technique in NSGA Ⅱ. The advantage of the algorithm over NSGA Ⅱ in terms of crowding mechanism was analyzed. In the experiments, dynamic model parameters identification problem of the methanol-to-hydrocarbons process was changed into minimum optimization problem by fitting the sampling data. The result demonstrated the effectiveness of the algorithm for dynamic model parameters identification, compared with the conventional methods.