Abstract: It is important to achieve the environmental, economic, and secure operation of power systems by improving active power flow distribution and reducing atmospheric pollutant and CO₂ emissions of thermal generators, which are achieved via optimal dispatch. Focusing on power systems with multiple types of generators, such as carbon capture power plants, wind generation, and conventional thermal generators, this paper considers the multiple factors such as the CO₂ and atmospheric pollutant emissions, the stochastic wind power, and the N−1 contingency. It proposes the stochastic active power dispatch model for the environmental, secure, and economic operation of power systems. In this model, the environmental and fuel costs of thermal generators, wind power costs, and the post-contingency correction control cost are included in the objective function. In the constraints of the proposed model, the constraints of normal operation and the secure constraints with post-contingency correction control are considered. Considering the characteristics of the proposed stochastic active power dispatch model, this paper proposes a fast and high-efficient method based on the fully connected deep neural network (FCDNN) to solve this model. In this method, FCDNN is used to obtain the initial point adopted in the optimization software, which can speed up the solving process. Finally, three modified IEEE test systems are used to validate the effectiveness of the proposed model and method.