研究报告

  • 郑霞,胡东滨,李权.基于小波分解和SVM的大气污染物浓度预测模型研究[J].环境科学学报,2020,40(8):2962-2969

  • 基于小波分解和SVM的大气污染物浓度预测模型研究
  • Study on prediction model of atmospheric pollutant concentration based on wavelet decomposition and SVM
  • 基金项目:新时代矿产资源开发与生态保护协调发展的理论与实证研究(No.71991483);中国工程院咨询研究项目(No.2019-XY-037);中南大学中央高校基本科研业务费专项资金资助项目(No.1053320184048)
  • 作者
  • 单位
  • 郑霞
  • 中南大学 商学院, 长沙 410083
  • 胡东滨
  • 1. 中南大学 商学院, 长沙 410083;2. 湖南省两型社会与生态文明协同创新中心, 长沙 410083
  • 李权
  • 中国地质大学(武汉) 环境学院, 武汉 430074
  • 摘要:针对大气污染物浓度的精准预测问题,运用小波分解将污染物浓度一维序列分解为高维信息,结合气象及污染物浓度数据,构建了基于小波分解的支持向量机预测模型.最后将模型应用于长沙市2018年PM2.5和O3-8 h的浓度预测.结果表明:①在其他参数不变的条件下,该模型在平均绝对误差(MAE)、平均绝对百分比误差(MAPE)、均方根误差(RMSE)、一致性水平(IA)和相关系数(R)指标上均优于未经小波分解的预测模型;②在考虑其他污染物对PM2.5浓度的影响后,预测模型评价指标MAE、MAPE和RMSE分别减少了5.57%、9.91%和3.44%,有着更小的误差;③在考虑气象因素对O3-8 h浓度的影响后,预测模型评价指标MAE、MAPE和RMSE分别减少了1.59%、3.54%和0.82%,同样也有更小的误差.由此可以看出,本文所提模型能够有效预测大气污染物浓度,为相关研究提供了方法参考.
  • Abstract:Responding to the problem of accurate prediction of air pollutant concentrations, this paper decomposes the one-dimensional sequence of pollutant concentrations into high-dimensional information based on wavelet decomposition and creates a support vector machine (SVM) prediction model on the basis of meteorological and pollutant concentrations data. The model is applied to predict the PM2.5 and O3-8 h concentration in Changsha in 2018. It is concluded that: ① With the presupposition that other parameters are fixed, indicators including mean absolute error (MAE), mean absolute percentage error (MAPE), root mean square error (RMSE), consistency level (IA), and correlation coefficient (R) in this predication model all perform better than that in predication model without wavelet decomposition.② When the impact of other pollutants on PM2.5 concentration is taken into account, the indicators of MAE, MAPE and RMSE are cut by 5.57%, 9.91% and 3.44% respectively, i.e. with much smaller errors.③ After considering the impact of meteorological factors on O3-8 h concentration, indicators of MAE, MAPE, and RMSE are reduced by 1.59%, 3.54%, and 0.82% respectively, also with smaller errors. It can be seen that the prediction model proposed in this paper could effectively predict the air pollutants concentration. As a result, it provides important reference for future related researches.

  • 摘要点击次数: 337 全文下载次数: 349