基于机器学习的PM_(2.5) 和O_(3)浓度预测模型研究  

Prediction model of PM_(2.5)and O_(3)concentration based on machine learning

作　　者：王宁[1] 李亚滨[2] 曾燕[3] 王勇[1] 朱晓晨[1] WANG Ning;LI Ya-Bin;ZENG Yan;WANG Yong;ZHU Xiao-chen(School of Ecology and Applied Meteorology,Nanjing University of Information Science&Technology,Nanjing 210044,China;The Climate Center of Heilongjiang Province,Heilongjiang Climate Change Center,Harbin 150030,China;Key Laboratory of Transportation Meteorology of China Meteorological Administration,Nanjing Joint Institute for Atmospheric Sciences,Nanjing 210041,China)

机构地区：[1]南京信息工程大学生态与应用气象学院,江苏南京210044 [2]黑龙江省气候中心黑龙江省气候变化中心,黑龙江哈尔滨150030 [3]南京气象科技创新研究院中国气象局交通气象重点开放实验室,江苏南京210041

出　　处：《环境生态学》2024年第11期83-93,98,共12页Environmental Ecology

基　　金：中国气象服务协会气象科技创新平台项目(CMSA2023MC022)资助。

摘　　要：本研究以江苏沿海地区为例,基于5种机器学习方法,利用2016-2019年的气象和环境监测站点数据以及地理信息、社会经济等开放数据建立了PM_(2.5)和O_(3)的浓度预测模型,通过SHAP值对模型进行可解释性分析并对空气质量做出评价。结果表明,极度梯度提升树(XGBOOST)模型对PM_(2.5)和O_(3)浓度的预测效果最好;气温、风速和相对湿度对PM_(2.5)浓度的预测影响最大,气温、日照时数、相对湿度对O_(3)浓度的预测影响最大;江苏沿海地区在2016-2019年的空气质量较好,PM_(2.5)污染得到了有效控制,O_(3)污染呈现波动变化,整体上有加重的趋势。This study takes the coastal areas of Jiangsu Province as an example and establishes prediction models for PM_(2.5)and O_(3)concentrations using five machine learning methods based on meteorological and environmental monitoring station data from 2016 to 2019,as well as geographical information,socioeconomic data,and other open data sources.The models are further analyzed for interpretability using SHAP values,and an assessment of air quality is conducted.The results indicate that the extreme gradient boosting tree(XGBOOST)model performs best in predicting PM_(2.5)and O_(3)concentrations.Temperature,wind speed,and relative humidity have the greatest predictive impact on PM_(2.5)concentrations,while temperature,sunshine hours,and relative humidity have the greatest predictive impact on O_(3)concentrations.The air quality in the coastal areas of Jiangsu Province was relatively good from 2016 to 2019,with effective control of PM_(2.5)pollution,while O_(3)pollution exhibited fluctuating changes,showing an overall worsening trend.

关 键 词：PM_(2.5) O_(3) 机器学习 SHAP 

分 类 号：X513[环境科学与工程—环境工程] X515