Wind power is an important clean and renewable energy. Integrating it into smart grid is significant to the energy conversion and emission reduction. In order to reduce the negative impact introduced by uncertainties and randomness of large scale wind power integration, a stochastic model predictive control (SMPC) based optimization and scheduling approach is proposed to coordinate to the wind power and traditional fossil generators. The discrete generation regulation constraints of some traditional generators without the automatic generation control(AGC) function are considered, and schedulable loads are introduced to make the system more flexible. A mixed integer quadratic programming (MIQP) based energy management model is constructed, and the regulation frequency constraints, minimum up/down time constraints and discrete output constraints are all considered. A two-stage scenario cutting method is proposed to efficiently choose typical scenarios. Experimental results show that the approach proposed is flexible and efficient by comparing with the traditional scheduling approach. Furthermore, the impact of start-up/shut-down times and schedulable loads is discussed.