学 术 报 告

报告题目：A new idea of predicting particulate matter 2.5(PM2.5) -- the perspective of data-driven PDE
报告人：张书华 教授  天津财经大学
报告地点：理学院五楼数学研究中心多媒体报告厅
报告时间：2020/11/01  16：00-16：45

数学与统计学院

2020.11.2

报告摘要: The real-time prediction and monitoring of PM2.5 is an important prerequisite for the prevention and control of air pollution. Although there are currently a variety of prediction methods, most of them are based on data rather than mechanism, so the prediction results cannot be explained by mechanism.  In addition, existing research methods are mostly applicable to local geographical areas, so it is impossible to study PM2.5 pollution from the perspective of transboundary pollution. Our previous work shows that partial differential equation (PDE) can predict PM2.5 in the short term. In view of this, we put forward two kinds of prediction ideas, which are based on data and mathematical equations: (1) we use data, networks, machine learning and the law of conservation of power system, based on the mechanism from the perspective of transboundary pollution of PM2.5, and finally propose PDE models of non-stationary condition. The PDEs can quantitatively depict the space-time characteristics of PM2.5 pollution and make short-term prediction; (2) based on the short-term historical data of PM2.5, we proposed the PDE (or ODE) construction method from the perspective of inverse problem by using stochastic optimization methods such as genetic algorithm and gradient method.