区域大气重污染的多目标非线性优化减排控制

This project studies the regional air heavy pollution control by emission reduction based on multi-objective non-linear optimization. First, focusing on the heavy haze events in the Beijing-Tianjin-Hebei region, to minimize both the PM2.5 concentration and the economic cost of emission reduction, selecting the control variables and constructing the objective functions, the emission reduction control can be coupled with the air quality model, and the multi-objective optimization is used to solve the problem; finally, the simultaneously control of complex air pollution is discussed. The core objective of this project is to provide an optimal emission reduction control strategy 3 days before the approaching heavy haze event is forecasted, reducing the impacts of heavy pollution on public health and the environment with the minimal economic cost. The feature and innovation point of this project are: (1) full consideration of non-linear characteristics and multi-objective optimization, its solving employ comprehensive disciplines such as atmospheric sciences, optimization control, and machine learning; (2) using the idea of big data, one “quasi-offline” method is developed to emergency response the heavy air pollution. After classifying the historical heavy haze events and identifying the current one as one type, the emergency “online” optimal control problem can be converted to the “offline” one which can be pre-solved for the whole type. Then the emission reduction strategy for the whole type is evaluated and adjusted “online” to find the final optimal control strategy for the current heavy haze event. This project is helpful for improving the emergency response level of heavy air pollution weather, providing scientific basis for regulating the mission reduction measures and decision-making, and it is also of great significance for the study of simultaneously control of complex air pollution.

本项目研究区域大气重污染的多目标非线性优化减排控制。首先针对京津冀地区的重霾天气，以PM2.5浓度和减排经济代价为控制目标，选择被控量，构建目标函数，与控制耦合进行多目标优化求解；最后探讨复合污染物的协同控制。项目核心目标是对预报的临近重霾天气，可在3天前提供优化减排的控制策略，降低重污染天气对公众健康和环境影响的同时，经济代价最小。特色与创新是：①充分考虑非线性特点和多目标优化，其解决综合了大气科学、优化控制和机器学习等多学科；②利用大数据的思想，发展了一套“准线下”方法及时应对大气重污染，对历史资料的重霾天气分类，将当前重霾天气归类，将此要求时效性的“线上”问题转换为可“线下”处理的、对该类型重霾天气的优化减排控制问题，再进行“线上”评估与调整。本项目有助于提高对重污染天气的应急处理水平，为“靶向施措、科学应对”制定减排措施与决策提供科学依据，也对复合污染物的协同控制研究有重要意义。

我国大气污染防治工作已进入攻坚期，“深入打好蓝天保卫战”要求精准科学施策，提高污染治理的针对性、科学性、有效性。对污染源进行减排控制是实现短期空气质量目标的有效手段，但大气重污染的减排控制是典型的多目标优化问题，企业限产、交通限行等减排措施必然造成经济损失和居民负担等，如何实现抓主要矛盾和平衡多目标的科学施策是亟待解决的科学问题与时代任务。本项目的研究任务即聚焦于找到对目标区域空气质量影响大的污染源及其合理、经济的减排方案，在达到空气质量目标的同时，尽量降低经济代价。.本项目的主要研究内容包含：大气污染优化减排控制的数学理论研究、大气重污染多目标优化减排控制方案与建模、大气重污染多目标优化减排控制求解方法和模拟示例。通过本项目的研究，取得了以下主要进展和成果：(1)理论分析了大气污染优化减排控制的数学模型和优化解的理想形式；（2）提出将当前重污染天气过程的优化减排控制这一实时“线上”问题转换为“准线下”计算的动态优化减排控制方案，解决了对当前重污染天气进行优化减排控制的时效性问题；(3)构建了一套包含关键被控变量、目标函数和约束条件的多目标优化减排控制模型，发展了基于进化算法的求解方法，得到同时优化PM2.5浓度指标和经济代价指标的Pareto最优解集和减排决策集，分析了PM2.5浓度指标和经济代价指标之间的博弈关系，给出了减排控制方案的高效率区和低效率区，得到的多目标优化减排方案可在较低经济代价下有效实现污染物的“削峰降速”。在示例模拟中，将空气质量目标选为本次重污染天气过程大于250μg/m3的PM2.5浓度均值，不加控制时该值高达303μg/m3，优化控制后降到258.57μg/m3，对比最高控制代价的结果，优化控制用最高代价的21.54%已能实现73.93%的目标。本项目研究有助于提高对重污染天气的应急处理水平，为制定精准科学的减排措施与决策提供科学依据和支撑。