The requirements for service quality and reliability of the electric power communication business are increasingly high. Based on software-defined network, a power communication network routing optimization strategy using the minimum path selection degree is proposed. The path selection degree is obtained by combining the prediction results of the link bandwidth occupancy rate prediction model based on the graph convolutional network with the triangular modulus operator. This is then used as the basis for the selection of the power service transmission path, and the dynamic allocation of the link resources of the power communication network is realized. Through simulation and traditional routing optimization strategy, the packet loss rate, transmission delay and link bandwidth occupancy rate are compared and analyzed to verify the superiority of the proposed strategy. Simulation results show that this strategy not only can meet the requirements of power transmission, but also link load distribution is more even, and it is better than traditional routing optimization strategies in terms of packet loss rate, transmission delay and link utilization. This work is supported by the Natural Science Foundation of Hebei Province (No. F2015502047).