Large oil-immersed power transformer fault diagnosis is always directed at preventive test data, the preventive test data can not be got immediately, otherwise, it must be waited until power off and maintenance time. On the other hand, the accuracy and quantity for the field data collection is always limited. However, the symptoms phenomenon is a summary of a great deal experiences, to some extent, it can reflect the failure of transformer. Therefore, an idea that integrates both of the preventive test data and the symptom phenomenon in the transformer fault diagnosis is proposed. Through this method, all kinds of information can complement each other. First, the diagnosis to symptoms phenomenon is realized by introducing fuzzy multi-attribute decision making (FMADM) theory. Then, by adopting the fuzzy probability model, the failure probability of the preventive tests data is calculated. Finally, through D-S evidence theory, the results of the preventive test data and the symptom phenomenon can be integrated. The paper gives a novel diagnosis model which can be used as a kind of effective means through the given example.〖JP〗