The state-of-the-art MW-frequency control is performed by two hierarchical mechanisms: economic dispatch (ED) and automatic generation control (AGC). The ED is solved every 5 min identifying the most economic generation dispatch and reserve schedule. AGC is a feedback control system that regulates area control error by sending signals to regulation reserve every 2-6 s. In system with high renewable penetration and associated high net-load variability, the conventional ED-AGC hierarchical model may result in degraded MW-frequency performance and increasing operational cost. In this paper, an AGC dynamic constrained ED model is proposed to provide a more reliable and economical regulation reserve schedule in handling high net load variability. The continuous AGC dynamics is transformed into discrete state-space model, and the fast AGC dynamics are eliminated to reduce the system order. The discretized reduced-order AGC dynamics are incorporated into the ED to optimize the regulation schedule. The proposed model can be applied in a look-ahead mode with very short-term load forecast or in a look-back mode to provide a "perfect dispatch" benchmark based on historical net load data. The 5-bus and 118-bus systems are tested to demonstrate the effectiveness of the proposed model.