This paper designs an interactive teaching-learning optimization(ITLO) algorithm for voltage source converter based high-voltage direct-current(VSC-HVDC) systems, which is used to optimize the control gains of proportional-integral(PI) control loops. Firetly, a wider exploration is achieved by introducing multiple classes into the teaching-learning based optimization(TLBO) algorithm. Then, a small world network(SWN) is employed for a deep interactive learning among the teachers or students from different classes, so that a more accurate exploitation can be realized. As a result, ITLO is able to effectively avoid a local optimum thanks to its proper trade-off between explorations and exploitations. Three case studies are undertaken, such as active and reactive power tracking, short-circuit fault at power grid, and wind farm integration. Simulation results show that the proposed approach has great advantage compared with typical meta-heuristic optimization algorithms.