A large number of harmonic currents are generated in the operation of converters in high voltage direct current (HVDC) systems. In order to reduce its adverse impact on AC systems and ensure the effective transmission of power in HVDC transmission lines, AC filters must be installed on AC buses. Aiming at the limitations of current design methods for AC filters in HVDC systems, this paper established the optimization mathematical model of AC filters, which aimed at the cost of filters and the total harmonic distortion rate of AC bus voltage and constrained the telephone harmonic form factor, and proposed an improved multi-objective particle swarm optimization (MOPSO) algorithm to achieve the multi-objective optimal design of AC filters. This algorithm used the second level repository of adaptive grid, and also added constraint processing mechanism and a special mutation operator to improve the exploration ability of the algorithm. The fuzzy decision method was used to select the optimal solution of multi-objective optimal design. The simulation model of ±500 kV HVDC system was established. The simulation results show that the AC filters designed by multi-objective optimization can effectively filter harmonics generated by the converter and save costs, and achieve the optimal balance between filtering performance and investment cost.