The economic dispatch problem (EDP) is a basic problem of energy management in the actual smart grids operation. This paper studies a general EDP of minimizing a sum of local convex cost functions subjected to both local interval constraints and coupling linear constraint over an undirected network. To reduce the amount of computation and interaction, we bring out a fresh event-triggered distributed accelerated primal-dual algorithm, named as ET-DAPDA, with uncoordinated step-sizes for solving the EDP. ET-DAPDA (with respect to the dual updates) incorporates two types of momentum terms into gradient tracking scheme and assumes that each node independently interacts with neighbors only at the event-triggered sampling time instants. Presuming the smoothness and strong convexity cost functions, the linear convergence of ET-DAPDA is analyzed by using the generalized small gain theorem. Moreover, ET-DAPDA strictly excludes the behavior of Zeno-like, thus heavily reducing the interaction cost. ET-DAPDA is studied on 14 bus and 118 bus systems to evaluate its applicabilities. The simulation results of the convergence rate are further compared with the prior art, which proves the superiority of ET-DAPDA.