For the failure of smart meter due to various external factors,this paper presents a classification and prediction model of smart meter fault type based on Bayesian networks. First of all, we analyzed the various factors that caused the failure of smart meter,and then training model with a large number of historical failure data, combined with expert advice, the Bayesian network structure is constructed based on scoring, The fault type prediction and decision analysis are carried out, and theSperformance ofStheSproposedSmethodisSverifiedS. The results show that this method can effectively predict the fault type of smart meter, and has high computational efficiency and good applicability.