Smart meter are widely used owing to the convenience of information collection and the perfection of their functions. How to maintain such a large number of smart meters efficiently and pertinently is a challenge for power operation enterprises. In order to solve this problem, a fault early-warning method of smart meter based on data mining technology is proposed in this paper. The fault early-warning model of smart meter is constructed by using c5.0 algorithm, and the model is trained through a large number of training sets, and then the early-warning accuracy of the model is obtained by using the test sets. A fault early-warning system is built through VS 2016 platform. The simulation results show that the system can accurately warn the running state of smart meter. According to the early-warning results, the electric power operation enterprises can carry on the key inspection to the abnormal meter according to the early-warning result, thus saving the manpower and material resources wasted due to the household investigation.