The AC filter in converter station in switching on will cause different levels of transient impulse.The phase-selection closing device can keep the three phases close near the voltage zero through indepen-dent phase operation of circuit breaker

and the transient impulse occurred is minimum. The reliability onselecting the setting of phase-selection closing device will affect the closing result. It is found through statis-tics that the mechanical wear due to long term operation of circuit breaker and variation of both mechanicaland electrical performance of each component due to aging

the actual closing will vary with the trends overtime

eventually causing the actual closing time deviate from the setting. In this paper

a kind of adaptiveselection method of phase selection closing setting by the use of PSO-based grey neural network is pro-posed. The mode is trained by the use of historical closing time together with PSO optimization algorithm topredict next closing time. The feasibility and efficiency of the method is proved by the prediction result.