Particle swarm optimization (PSO) develops rapidly for its simple code and easy operationThe performance of PSO rests with two critical factors: inertia weight and acceleration coefficients A proper PSO configured inertia weight and acceleration coefficients are presented and applied to the variational data assimilation (VDA) In order to verify its effectiveness, a simplified partial differential equation containing discontinuous “on off” switch is used as the governing equation and three kind of comparative assimilation numerical experiments (VDA based on the conventional adjoint method, genetic algorithm and PSO) are conducted, respectivelyThe numerical results show that the quality of VDA with “on off” switches based on PSO is much better than the one based on the other two algorithms, and the performance of PSO during the optimization is most stable Moreover, the sensitivity experiment for observational noise and model errors shows that PSO possesses more strong robust characteristics comparing to the conventional adjoint method and genetic algorithm In addition, it is shown that the effectiveness of VDA based on PSO is related to the configuration of algorithm parameters, more proper parameters resulting in higher quality of assimilation results.