基于对电力公司历史发电及排放数据(二氧化硫、氮氧化物和烟尘)的分析,文中首先采用递归神经网络(RNN)拟合出发电功率与排放数据回归模型,然后在该回归模型与蚁群算法(ACO)的基础上设计实时发电调度算法,在满足实时发电任务的前提下使机组总污染物排放量降低,达到节能减排的目的.文章最后通过安徽省电力的真实数据构建仿真实验验证回归模型和调度方案的有效性. Based on the analysis of power company historical generation and emission data (sulfur dioxide, nitrogen oxides and soot), a regression model of power generation and emission data was first fitted by using recurrent neural network (RNN). Then, the regression model and ant colony algorithm ACO) based on the design of real-time generation scheduling algorithm, to meet the real-time power generation tasks under the premise of reducing the total unit emissions of pollutants to achieve energy-saving emission reduction purposes.At the end of the real data through the Anhui Provincial Electric Power Construction Simulation Experimental Verification Regression Model And scheduling the effectiveness of the program.