The fault tree of the relay protection system is an important tool for risk identification and evaluation of the system. To solve the difficulty in precisely obtaining the toplevel event rate of fault tree, and the inability of reverse reasoning of fault trees, a fault risk assessment method for relay protection which is based on the fault tree and the Bayesian network is proposed. In the positive direction, the fault state of the root nodes in the Bayesian network (corresponding to the basic event of the fault tree) can be confirmed according to the realtime operational data of relay protection. Combined with by the Bayesian network reasoning, the fault probability of the leaf node of Bayesian network (corresponding to the top event of the fault tree) can then be generated, thus to realize the risk prediction of the protection system. In the reserve direction, the leaf nodes of Bayesian network are combined with the Bayesian conditional probability formula, to calculate the failure probability of the root nodes of the Bayesian network, therefore the positioning and tracing of the fault cause can be achieved. The proposed method can provide bases for the reliability improvement and the fault diagnosis of the protection system.