With the continuous development of power Internet of Things, the network has gradually become a bridge between intelligent power terminal equipment and intelligent computing communication. The data generated by power equipment will be processed through the network for intelligent computing. How to better balance the scheduling nodes and achieve the optimal scheduling routing mechanism under the limited arithmetic resources has become an urgent problem to be solved. To this end, this paper first builds an arithmetic-aware power dispatch network (AAPDN) architecture, which adopts a distributed routing control scheme to avoid the situation where large-scale power data are processed, parsed, stored and transmitted in a single centralised point, which leads to a single point of failure caused by an overburdened computational routing. On this basis, this paper further introduces an arithmetic routing notification and decision-making mechanism, which is used to share the arithmetic state information of the scheduling service nodes and formulate the optimal routing and scheduling decisions. Simulation experiments show that the mechanism proposed in this paper can maintain a high level of power scheduling service satisfaction while ensuring the arithmetic usage rate.