In the smart distribution gird, various communication technologies are adopted to form a seamless heterogeneous communication network to deliver control and protection signals. However, most scheduling strategies designed for smart gird aim at optimizing system operation performance without considering the transmission cost. To address this problem, we construct a delay-constrainted cost optimization model to accomplish the following two goals: to optimize the cost and to satisfy QoS requirements. Firstly, we establish a queuing model according to the characteristics of heterogeneous services and output networks for smart grid. Then, a delay-constrainted optimal traffic allocation strategy is designed to dynamically allocate heterogeneous service data to different output networks. The allocation strategy herein is based on the Lyapunov theory. Finally, simulation results show that our proposed allocation strategy significantly reduces the cost and meets transmission delay constraints for all service traffic.