In order to overcome the limitation of the traditional expert system applied in fault diagnosis of power system

a query method for optimal diagnosis of power system faults is proposed

which is based on model-based diagnosis. With the diagnosis acquiring method based on the causal relation

the preparation candidate diagnoses are obtained according to presupposed fault output. Then the candidate diagnoses are determined by the matched presupposed fault output according to electric data after faults occurring. Finally

the fault probabilities of candidate diagnoses are obtained by applying the Bayesian theory. Based on those probabilities

the optimal diagnosis is queried from the candidate diagnoses. The goal of reducing the diagnosis time is achieved by getting ready candidate diagnosis offline and getting candidate diagnostic online

and in order to improve the diagnosis accuracy

alarm data are introduced to logic framework of model-based diagnosis for computing the actual probability of fault about power system element. The simulation results show that the proposed method is effective and practical.