1: \begin{abstract}

2: A consensus-control-based multi-level control law named \emph{Multi-Level Power-Imbalance Allocation Control} (MLPIAC) is presented for a large-scale power 

3: system partitioned into two or more areas. 

4: Centralized control is implemented in each area while distributed control is 

5: implemented at the coordination level of the areas. Besides

6: restoring nominal frequency with a minimal control cost,

7: MLPIAC can improve the transient performance 

8: of the system through an accelerated convergence of the control inputs without

9: oscillations. At the coordination level of the control areas, 

10: because the number of the areas is smaller 

11: than that of nodes, MLPIAC is more effective to obtain the minimized 

12: control cost than the purely distributed control law. 

13: At the level of the control in each area, because the number of nodes 

14: is much smaller than the total number of nodes in the whole network,

15: the overheads in the communications and the computations are reduced compared to the 

16: pure centralized control. The asymptotic 

17: stability of MLPIAC is proven using the Lyapunov method and 

18: the performance is evaluated through simulations. 

19: \end{abstract}

20: