With the gradual commercialization of 5G, the construction of the data center ushers in explosive expansion, which leads to higher energy consumption and is not conducive to energy protection, emission reduction and sustainable development. At the same time, the "dual carbon" goal is proposed, which requires green and low-carbon development of all industries. Therefore, this paper proposed an energy-saving scheme integrated frequency conversion and Artificial Intelligence (AI) to obtain higher energy-saving efficiency for Internet Data Center (IDC). Besides, the cloud-edge-end architecture is utilized to collect sufficient multi-type data for model training. Such a structure can help generates an overall strategy, and finely adjust the fan threshold of the frequency converter effectively solving the problems of local optimization and extensive parameter setting in the frequency conversion, The experiment conducted in the air-cooled constant frequency IDC room verifies that the energy saving rate of the proposed solution can reach 34.6%.