Abstract: The optimal dispatch of distributed generators relies on the communication between devices. The communication delay affects the dispatching process and may even cause the instability of the system operation. Aiming at the economic dispatch problem of the multiple distributed generators in the case of a sparse communication network with communication delays, a fully distributed economic dispatch strategy based on the diffusion algorithm is proposed. First, a multi-agent structure for controlling the distributed generators is designed so that the communication and control between these multiple agents do not depend on any "leader" unit. An economic dispatch model of the system is established, which aims at the lowest operating cost of the whole system. The active power of the distributed generators and the demand-side flexible loads is optimized. On this basis, the influence of communication delay on the active power dispatch is studied, and an improved diffusion strategy supporting fully distributed optimization is designed. Compared with the consensus method, the improved diffusion strategy can ensure better performance under the same delay with faster convergence speed and higher stability. Simulation results of several scenarios verify the correctness and effectiveness of the proposed strategy.