Abstract: Based on platform of three-dimensional particle-in-cell (PIC), CHIPIC, modules of particle swarm optimization (PSO) and genetic algorithm (GA) are designed to optimize a relativistic backward wave oscillator (RBWO), respectively. Comparisons of PSO and GA are implemented in three kinds of parameters of RBWO:Continuous parameter, discrete parameter, and mix parameters. It shows that performances of PSO are better than that of GA. PSO has higher optimization accuracy and convergence rate than GA.