Air pollutants generated by thermal power plants have been a major source of environmental and health hazard. This paper develops an optimal mechanism for generation scheduling of a hybrid power system consisting of conventional and renewable power plants while maintaining the air-pollutant concentration below a targeted level. To capture the physical movement of the air pollutants, the proposed framework applies a two-dimensional advection-diffusion model and discretizes it into a discrete-Time state-space model. Due to the limited sensing, we design a Kalman filter for data assimilation. Based on the proposed mechanism, the integration of the Independent System Operator (ISO) with the sensors and power plants constitutes a feedback system for the efficient operation of the smart energy systems. To accelerate the computations, we decompose the original problem into small sub-problems and design a decentralized algorithm to provide solutions to the environmentally constrained power dispatch problem. We use case studies to evaluate the influence of the pollution constraints on the solutions and the impact of the wind speed on the pollution constraints.