It's difficult to obtain the analytic solution of complicated optimization problems in engineering. To try to get numerical solution with numerical computation method is easy to get local optimal solution instead. In order to improve the speed and accuracy of numerical optimal solution to high-dimensional problems, the population is divided into intelligent population and common population on the basis of traditional culture algorithm, and they are initially optimized according to different evolving ways respectively. Furthermore, simulation of practical examples by MATLAB program is carried out. The results show that the speed is greatly improved after optimization and the cultural algorithm with initialized improvement can approximate the optimal solution faster and more accurately in high-dimensional optimization problems. The optimized cultural algorithm provides a new reference for practical engineering problem.