基于奇异谱分析的多模型融合空气污染物质量浓度预测方法  

Prediction Method of Air Pollutant Concentration Based on Singular Spectrum Analysis and Model Fusion

作　　者：徐洪学[1] 孙万有 杜英魁[1] 汪安祺 王阳阳 李娟[2] XU Hongxue;SUN Wanyou;DU Yingkui;WANG Anqi;WANG Yangyang;LI Juan(College of Information Engineering,Shenyang University,Shenyang 110044,China;College of Physical Education,Shenyang University,Shenyang 110044,China)

机构地区：[1]沈阳大学信息工程学院,辽宁沈阳110044 [2]沈阳大学体育学院,辽宁沈阳110044

出　　处：《沈阳大学学报（自然科学版）》2021年第6期470-479,共10页Journal of Shenyang University：Natural Science

基　　金：国家自然科学基金资助项目(61873338);国家社会科学基金资助项目(19BTY053);辽宁省重点研发计划(2019JH2/10300014)。

摘　　要：提出一种基于奇异谱分析的多模型融合空气污染物质量浓度预测方法,它基于奇异谱分析的时间序列分解特性,对差分整合移动平均自回归模型和灰度模型进行模型融合。采用模型融合方法对沈阳市京沈街区空气污染物质量浓度监测站点2020年9月18日—2021年3月11日空气污染物质量浓度监测数据进行预测分析。结果表明,相较于单独使用差分整合移动平均自回归模型方法或灰度模型方法,此方法能够显著提高空气污染物质量浓度的预测精准度,对空气污染物质量浓度变化能够达到较好的预测效果。A multi-model fusion air pollutant concentration prediction method based on singular spectrum analysis was proposed.Based on the time series decomposition characteristics of singular spectrum analysis,the differential integrated moving average autoregressive model and gray scale model are model-fused.The model fusion method was used to predict and analyze the air pollutant concentration monitoring data of the air pollutant concentration monitoring site in the Jingshen District of Shenyang City from September 18,2020 to March 11,2021.The results show that compared with the use of differential integrated mobile alone average autoregressive model method or gray-scale model method,the proposed method can significantly improve the accuracy of air pollutant concentration prediction,and achieve a better prediction effect on air pollutant concentration changes.

关 键 词：空气污染物 预测 奇异谱分析 时间序列 灰度模型 自回归模型 

分 类 号：TP399[自动化与计算机技术—计算机应用技术]