A feature extraction method of principal component analysis (PCA)was introduced. By the application of dimension reduction, the decision variables were mapped from the highdimensional to the low dimensional one, and the principal components (main decision variables) were obtained. The simplified model with improve efficiency combined with Extreme Learning Machine (ELM) was constructed. The multiobjective optimization model was established by using of an improved strategy of nondominated sorting genetic algorithm (NSGA II). Pareto optimal solution sets were solved.