With the large-scale interconnection of Renewable Energy Sources (RESs), their intermittent and stochastic characteristics bring great challenges to the frequency stability and control of the power system. Based on the existing automatic generation control framework, an optimal control framework based on model predictive control (MPC) is proposed according to the complementary characteristics of conventional generators and plug-in electric vehicles (PEVs). The MPC load distributor can control the regulation power from conventional generator sets and PEVs according to the dynamic and AGC execution cycles of conventional generator sets and the PEVs at different time scales, so that the ACE tracking error and the adjustment cost can be minimized by meeting regulation power and power constraints. Simulation results show that the proposed method can achieve AGC performance improvement and cost saving. This work is supported by National Natural Science Foundation of China (No. 50937002).