with the popularity of smart meters and the complexity of their application functions, it is more and more important to accurately predict the reliability of smart meters. Firstly, using the strong semantic characteristics of go graph, the architecture block diagram of electric energy meter is transformed into go graph; secondly, according to the Bayesian mapping rule of go method, go graph is transformed into Bayesian network to solve the problems of numerous operators and complex algorithm of go graph; and then, the conditional probability table (CPT) of each sub node of Bayesian network is annotated, and the sorted Bayesian network is transformed into coding Finally, the reliability of the meter and the posterior probability distribution of each node are obtained, and the fault of the meter is inferred. The accelerated life test of the meter is carried out, and the test results verify the scientificity and effectiveness of the method.