The particle swarm optimization(PSO) algorithm can be used to seek optimal load of the wave power generation system, but has low probability in searching global optimization. Therefore a novel artificial fish swarm algorithm was proposed to solve the problem. By comparing the object function value of the next position getting from the “chasing the trail behaviour or preying behaviour or random behaviour” and “swarming behaviour or preying behaviour or random behaviour”, the better behaviour mode was selected to confirm the search direction. The position of artificial fish was updated and the renewal object function value was calculated to solve the optimization problem by joining the random moving step. The simulation results show that the average output power of the wave power generation system is increased and the seeking of optimal load under different frequencies is achieved with the proposed algorithm. By comparing the traditional PSO, the novel algorithm can make the system effectively avoid the local optimization.