Relying on its technology and deployment advantages at the edge of the network, edge computing can collect the running status of terminal equipment in real time, and complete the upload of data to the cloud platform on demand, effectively alleviating the working pressure of the cloud processor and reducing the delay of real-time data return. However, the inherent characteristics of a single edge node with limited resources may still be unable to meet the delay requirements of large-scale access power Internet of Things services. In this regard, this paper studies the workload distribution that satisfies lower service latency. Considering the network delay and computing delay, a workload distribution model based on multiple edge computing nodes is constructed to minimize the service delay in the power Internet of Things scenario. In addition, in order to further reduce the service delay of multi-service terminals, a workload distribution mechanism has been developed. In the initialization phase, the task balancing algorithm is used to improve the workload balance among edge nodes. An improved particle swarm optimization algorithm is proposed to allocate the CPU resources of edge nodes. In task allocation, based on a semi-definite planning algorithm, application tasks are allocated to edge nodes to achieve a reasonable allocation of network resources. Finally, the simulation results verify the effectiveness of the task allocation strategy proposed in this paper.