The development of remote fault diagnosis of transmission line is one of the focuses of the construction of smart grid. To solve the problems of homogeneous uploaded data, enormous amount of data, and poor data utilization, naive Bayes algorithm is used together with similarity fault matching by time sequence to establish a transmission line fault warning model, based on the big data processing cluster technology provided by the background platform. The model uses collected data of electrical and switching information, event sequence information, power network topology, and data acquired by fault recorders during transmission line failure, and applies naive Bayes algorithm to excavate the occurrence factor of potential fault (i.e., fault factor), then, by cooperating with similarity fault matching by time sequence, the fault of transmission line can be forecasted. Case analysis shows that the model can excavate fault factors correctly and the forecast result is satisfactory.