This paper studies economic power dispatch strategy in a smart grid by using the distributed consensus protocol in multi-agent systems, where there are many generation units working cooperatively under a local neighboring area to achieve a minimum solution for the optimization problem of the total cost function. At first, the analysis indicates that the total cost for generators in a smart grid ... [Show full abstract] can reach its minimal value if the incremental cost for all generation units can achieve consensus and the balance between supply and demand of powers is satisfied. Then, by designing a distributed consensus protocol in multi-agent systems for the incremental cost with appropriate initial conditions, incremental cost consensus can be reached for all the generation units and the balance for the powers can also be kept. Thus, the optimization problem for the cost function of all generation units can be solved. Finally, some simulation examples are performed to verify the analysis in this paper.