The rapid development of digital economy promotes the explosive growth of data volume, which makes the construction of new infrastructure and its energy consumption emissions increase dramatically. Therefore, it is necessary to establish a scientific and objective energy consumption evaluation system and corresponding energy conservation measures. This paper first describes the building process of the energy consumption model, and then compares the difference between the traditional energy consumption model and the machine learning energy consumption model. It shows that it is an irresistible trend to use machine learning technology to build the energy consumption model. By applying the reinforcement learning model to the energy conservation of 5G base stations, it illustrates the actual application effect of the current network on reducing consumption.