AC filter in converter station is an important part of HVDC transmission system and the occurrence of tripping will affect directly the transmission power of the DC transmission system. In this paper

a kind of healthy status of AC filter which is on-line identified by the opening and closing current of AC filter circuit breaker is proposed. Firstly

both time and frequency domain feature of the opening and closing current of AC filter breaker is defined. On this basis

the fault warning of AC filter is achieved by the artificial intelligence method based on radial neural network. It is shown by the actual result of certain converter station that the method has high fault warning accuracy

can give warning before expansion or negative influence due to abnormality of the AC filter

and remind related people to check and deal with the defects on time so to reduce the occurrence of abnormal tripping of the AC filter.