The reliability test requirement for electric energy meter is more and more strong, the reliability prediction of intelligent electric energy meter is an effective means to evaluate the reliability of software. In this paper, the G-O model is used to evaluate the software analysis of intelligent electric energy meter software environment and extract test methods,and then the reliability of the electric energy meter is predicted from the number of test problems in the electric energy meter, and the products of different specifications can be compared, so as to deduce the stable period of the product. The research on reliability prediction of embedded software can guide the development and design of intelligent electric energy meter, which has certain reference value for improving the product quality and reliability.