An industrial Model Predictive Control method, which combines usual linear control and sequence control, is re-formulated to a hybrid control algorithm. The plant start up and shut down sequence is naturally used in the industrial control method, but has not been analyzed theoretically. In this paper, the Thermal Power Plant optimal load distribution problem is formulated as Model Predictive Control with the constrained condition, and also the embedded start up and shut down optimal control sequence is designed with different objective functions. The new hierarchical control system structure is discussed as a hybrid control system, which can dramatically reduce the calculation effort, in spite of plant non-linearity, constraints and parameter uncertainty. A simulation example applied to a thermal power plant is demonstrated to show the effectiveness of the control method.