Design of passive power filters shall meet the demand of harmonics suppression effect and economic target. However, existing optimization methods for this problem only take technology target into account or do not utilize knowledge enough, which limits the speed of convergence and the performance of solutions. To solve the problem, two objectives including minimum total harmonics distortion of current and minimum cost for equipments are constructed. In order to achieve the optimal solution effectively, a novel multi-objective optimization method, which adopts dual evolution structure in cultural algorithms, is adopted. Implicit knowledge describing the dominant space are extracted and utilized to induce the direction of evolution. Taken three-phase full wave controlled rectifier as harmonic source, simulation results show that filter designed by the proposed algorithm have better harmonics suppression effect and lower investment for equipments than filter designed by existing methods.