Abstract: Over the past few years, the backbone power grid of China have experienced a period of rapid expansion, the transmission channels of which cover a wide region of diversified terrain and complex environment, and it poses a huge challenge for ensuring the safe and stable operation of the important transmission channels. To solve the problems of fault identification and reliability evaluation of on-line monitoring devices over transmission line, various models of abnormal patterns were established based on in-depth analysis of abnormal data. Combining fault tree analysis and Bayesian network, a fault identification method for on-line monitoring device was proposed. Aiming at the shortcomings of the online device reliability evaluation method, a ternary judgment model including device state of failure, abnormality and normality was constructed. At last, the validity shows that this method can effectively determine the status of the on-line monitoring device and give the cause of failure and its occurrence probability, which can help the on-site operation and maintenance personnel to take targeted measures, which proved that this method has a broad prospect of engineering application.