Recent years, 5G networks have become an important role in accelerating the development of social intelligence. But it also increases energy consumption and data flow. In order to guarantee the experience of network users, an adaptive SWIPT-based MEC network is proposed. The network consists of multiple user equipment (UE) and multiple mobile edge computing (MEC) servers. The MEC server can make up for the shortcomings of the UE’s insufficient computing capability, and Simultaneous Wireless Information and Power Transfer (SWIPT) can send energy to the UE without pollution to make up for the shortcomings of limited battery energy. Added, we increase the utilization of sub-channels, improve the adaptability of the SWIPT-based MEC network to the environment, lengthen the battery life, and optimize the UE’s energy efficiency of the network. We also propose a three-part alternative optimization algorithm framework based on the categories of optimization variables. The first part combines the Alternating Direction Multiplier Method (ADMM) and Dinkelbach’s algorithm to optimize continuous variables. And for directly optimize the integer variables of the other two parts, without converting them to continuous variables. The second part adjusts the offloading decision by comparing the energy consumption of the two computation modes, and the third part proposes the integer Bat algorithm to assign sub-channel. The simulation results show that the energy efficiency of the binary offloading algorithm proposed in this article is 11% higher than the approximate algorithm based on the binary offloading KKT algorithm, and the relationship between various preset parameters and energy efficiency in the network is discussed.