A real-time distributed economic dispatch method for grid-connected microgrids is proposed in this paper to address the economic decline caused by network attacks. Firstly, a multi-agent-based discrete economic scheduling model is constructed, and a leading agent is established at the Public Connection Point (PCC). Secondly, PCC active power is tracked in real-time using consistency algorithms, and detection algorithms are designed to identify network attacks from neighboring distributed power sources. Subsequently, the attacked distributed power sources are gradually marginalized or isolated by adjusting the reputation weights in the communication adjacency matrix. The simulation results demonstrate that even if the network attacks some generator units, the optimal economic dispatch goal can still be achieved in a distributed manner by the remaining normal units. It is proved that this strategy can effectively resist common network attacks.