A large number of distributed energy resources (DERs) exist in the virtual power plant (VPP) system. To solve the VPP's economic dispatch problem, not only the power balance between supply and demand but also the power output constraints of all kinds of DERs should be considered. The distributed dispatch method usually has lower communication cost, easier implementation, higher flexibility and stronger robustness. In this paper, a distributed gradient algorithm is employed to solve this kind of economic dispatch problem. This method can handle both equality and inequality constraints effectively. The IEEE-34 bus test system is employed to demonstrate the validity of the proposed method. Experimental results verifies that the distributed gradient algorithm has the characteristics of fast convergence and high stability, can realize the online operation optimization in a distributed way, and manifest a superiority compared to centralized dispatch.