兰州大学机构库 >数学与统计学院
基于EMD的AQI和PM2.5预测建模研究
Alternative TitleResearch on AQI and PM2.5 Forecasting Based on EMD
练秀缘
Thesis Advisor朱素玲
2018-04-10
Degree Grantor兰州大学
Place of Conferral兰州
Degree Name硕士
Keyword空气污染 AQI EMD SVR GCA PSOGSA
Abstract

大气污染是影响公众健康、生态文明和社会可持续发展的重要因素,也是当今世界关注的热点。PM2.5对人体健康和大气环境质量会产生严重的影响,据世界卫生组织报道,世界上92%的人口生活在PM2.5水平超过世卫组织限值的地区。大气污染不仅严重影响着人类身心健康,而且严重破坏生态坏境,给人们的日常生活带来诸多不便。因此,建立一个可靠的空气污染预测预警系统是很有必要的,以便人们能够提前知道未来空气质量的好坏程度,并做好相应防护措施;为决策部门控制污染物排放提供参考。

AQI是衡量空气污染程度的重要综合指标, PM2.5是评价空气污染的最重要单项指标之一。根据国内外研究现状,目前很少有关于AQI的预测研究,并且结合时间序列模型和多元回归理论的PM2.5建模研究稀缺,本文结合空气污染预测较多利用时间序列建模的事实,主要开展以下两个方面的研究:AQI的时间序列建模研究。AQI的预测研究主要包括:基于EMD方法对原始AQI序列进行数据预处理;对分解得到的IMFs分别采取GM、Holt-Winter和S-ARIMA进行预测分析,进而建立了EMD-IMFs-Hybrid模型;对分解得到的IMFs之和采用SVR进行预测分析,进而建立了EMD-SVR-Hybrid模型。本文提出的EMD-SVR-Hybrid和 EMD-IMFs-Hybrid均利用S-ARIMA模型进行误差矫正。为了验证所提出理论的有效性,本文利用邢台市的AQI时间序列进行预测分析,对比模型为EMD-GRNN, Wavelet-GRNN, GRNN, SVR, Wavelet-SVR和ARIMA。从模型预测能力评价指标上看,EMD-SVR-Hybrid和 EMD-IMFs-Hybrid的预测误差(MAE、MAPE、RMSE)比六个对比模型低,与真实值的相关性指标IA和AI比对比模型高。因此,本文提出的AQI预测模型的预测精度较高、预测结果可靠。第二,PM2.5序列的混合建模研究。基于CEEMD、PSOGSA、SVR、GRNN和GCA方法建立CEEMD-PSOGSA-SVR-GRNN混合模型来预测PM2.5,其主要内容包括:使用CEEMD分解处理原始PM2.5序列进行数据预处理、对每个IMF分别采用PSOGSA-SVR做预测、对于PSOGSA-SVR预测的残余信息使用多元回归GRNN进行校正,其中GRNN影响因素利用GCA进行选择,基于以上步骤建立的模型称为CEEMD-PSOGSA-SVR-GRNN。为验证CEEMD-PSOGSA-SVR-GRNN的有效性,本文选用了三个城市(重庆、哈尔滨和济南)的PM2.5数据进行建模分析,对比模型为CEEMD-PSOGSA-SVR*, EEMD-PSOGSA-SVR, PSOGSA-SVR, CEEMD-PSO-SVR, CEEMD-GSA-SVR和CEEMD-GWO-SVR。结果表明,与其他对比模型相比较,CEEMD-PSOGSA-SVR-GRNN模型有较低的预测误差,也就是说提出的CEEMD-PSOGSA-SVR-GRNN模型有着较高的预测精度。本文所提出的预测方法,不仅可以用于AQI和PM2.5的预测分析,也可以应用于其他污染物(PM10, SO2, CO)的预测分析。总之,本文不仅弥补了空气污染统计预测理论研究的不足,同时为完善空气污染预测预警系统提供理论指导,而且该研究结果具有很好的应用价值。

Other Abstract

Air pollution is an important factor affecting public health, ecological civilization and social sustainable development,and it is also a hot topic in the world today. Especially PM2.5 has a serious impact on human health and the quality of the atmosphere. According to the World Health Organization (WHO), 92% of the world's population lives in areas where PM2.5 levels exceed the WHO limit. The air pollution not only seriously affects the physical and mental health of human beings, but also severely damages the ecological environment and brings many inconveniences to people's daily life. Therefore, it is necessary to establish a reliable air pollution forecast warning system, so that people can know the level of air quality in advance and take corresponding precautions. It also provides reference for decision-making departments to control pollutant emission.

AQI is an important comprehensive indicator to measure the degree of air pollution, and PM2.5 is one of the most important indicators to evaluate air pollution. According to the domestic and foreign research status, there are few studies on the prediction of AQI, and no PM2.5 hybrid modeling with time series model and multiple regression theory. The main aspects of our research are as follows:

 First, AQI forecasting. The AQI prediction research mainly includes:  Data pre-processing of the original AQI sequence based on the EMD; The decomposed IMFs are analyzed by GM, Holt-Winter and S-ARIMA respectively to establish the EMD-IMFs-Hybrid model; The sum of the decomposed IMFs are analyzed by SVR to establish the EMD-SVR-Hybrid model. In this research, the S-ARIMA model is employed to correct the residuals of EMD-SVR-Hybrid and EMD-IMFs-Hybrid. In order to verify the effectiveness of the proposed theories, this paper uses the AQI time series of Xingtai to carry out the prediction and analysis. The comparison models are EMD-GRNN, Wavelet-GRNN, GRNN, SVR, Wavelet-SVR and ARIMA. According to the evaluation indexes of model predictive ability, the forecasting errors (MAE, MAPE, RMSE) of the proposed model is lower than the six comparison models. Therefore, the AQI model proposed in this paper has high prediction accuracy and reliable forecasting results.Second, PM2.5 forecasting. Based on CEEMD, PSOGSA, SVR, GRNN and GCA methods, this paper establishes the CEEMD-PSOGSA-SVR-GRNN hybrid model to predict PM2.5 and its main contents include: Data pre-processing of the original PM2.5 sequences based on the CEEMD; For each IMF, we use the PSOGSA-SVR to make predictions; The residual information of the PSOGSA-SVR is corrected by multiple regression GRNN, and the inputs of GRNN are selected by GCA, that is called CEEMD-PSOGSA-SVR-GRNN. In order to verify the validity of CEEMD-PSOGSA-SVR-GRNN, three PM2.5 series of Chongqing, Harbin and Jinan are selected for forecasting comparisons. The comparison models are CEEMD-PSOGSA-SVR*, EEMD-PSOGSA-SVR, PSOGSA-SVR, CEEMD-PSO-SVR, CEEMD-GSA-SVR and CEEMD-GWO-SVR. Compared with other comparison models, the CEEMD-PSOGSA-SVR-GRNN model has a lower prediction error, which means that the proposed CEEMD-PSOGSA-SVR-GRNN model has higher forecasting accuracy.

The forecasting methods proposed in this paper can not only be used for prediction analysis of AQI and PM2.5, but also can be applied to the forecasting analysis of other pollutants (PM10, SO2, CO). Therefore, this research not only makes up the deficiency of the theoretical research on the statistical prediction of air pollution, but also has good application value.

URL查看原文
Language中文
Document Type学位论文
Identifierhttps://ir.lzu.edu.cn/handle/262010/225240
Collection数学与统计学院
Recommended Citation
GB/T 7714
练秀缘. 基于EMD的AQI和PM2.5预测建模研究[D]. 兰州. 兰州大学,2018.
Files in This Item:
There are no files associated with this item.
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Altmetrics Score
Google Scholar
Similar articles in Google Scholar
[练秀缘]'s Articles
Baidu academic
Similar articles in Baidu academic
[练秀缘]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[练秀缘]'s Articles
Terms of Use
No data!
Social Bookmark/Share
No comment.
Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.