基于CEEMDAN和Informer的上海市NO<sub>2</sub>浓度预测模型

尹祚成; 柴天

doi:10.13205/j.hjgc.202512022

基于CEEMDAN和Informer的上海市NO₂浓度预测模型

doi: 10.13205/j.hjgc.202512022

尹祚成,
柴天^,

厦门理工学院环境科学与工程学院, 福建厦门 361024

基金项目:

福建省产学研联合创新项目（2021Y4005）；福厦泉国家自主创新示范区城乡生态环境治理提升关键技术协同创新平台项目（2022-P-024）；福建省科技厅引导性项目（2020Y0056）

详细信息

作者简介:
尹祚成（1998—），男，硕士研究生，主要研究方向为基于深度学习的空气污染物预测。2576337938@qq.com

通讯作者:
柴天（1982—），男，副教授，硕士生导师，主要研究方向为环境系统工程和智能环境监测。chait@xmut.edu.cn

计量
- 文章访问数: 4
- HTML全文浏览量: 1
- PDF下载量: 0
- 被引次数: 0
出版历程
- 收稿日期: 2024-09-05
- 录用日期: 2024-11-02
- 修回日期: 2024-10-14
- 网络出版日期: 2026-01-09

A prediction model for air NO₂ concentration in Shanghai based on CEEMDAN and Informer

YIN Zuocheng,
CHAI Tian^,

School of Environmental Science and Engineering, Xiamen University of Technology, Xiamen 361024, China

摘要

摘要: 准确预测二氧化氮（NO₂）浓度对于预警空气污染风险、指导政策制定以及保护公共健康至关重要。为此，提出一种新的基于自适应噪声的完全集合经验模态分解（CEEMDAN）和Informer深度学习模型相结合的混合模型，用于预测上海市NO₂日均浓度。针对NO₂浓度时间序列的高噪声、波动性和非线性等特点，首先采用CEEMDAN对原始NO₂序列进行分解。在分解之前，引入灰狼优化算法（GWO），并将最小样本熵（SE）作为适应度函数，实现对CEEMDAN关键参数的自动寻优。由此，能够有效减少原始数据序列的随机波动，得到一系列本征模态函数（IMFs）。综合运用皮尔逊相关系数（Pearson）和斯皮尔曼相关系数（Spearman）分析多种污染物的关联强度，从中为各IMF筛选最佳特征，以构建高耦合度的特征矩阵，并将其分别导入Informer模型进行编码和建模。通过多头概率稀疏自注意力（Multi-head ProbSparse self-attention）和注意力蒸馏（Distilling）等机制，提高预测的精度和效率，最后将各个IMF的预测结果相加得到最终的预测值。通过选取多种基准模型进行对比，结果表明，所提模型的预测精度优于原始Informer模型以及其他基准模型，成功实现了预期效果。
- NO₂浓度预测 /
- 深度学习 /
- Informer /
- CEEMDAN /
- 特征选择
Abstract: Nitrogen dioxide （NO₂） is a significant pollutant in the troposphere， posing substantial threats to the environment and human health. Accurately predicting NO₂ concentrations is crucial for air pollution management， policy formulation， and public health protection. This study focused on Shanghai as the research area and proposed a novel hybrid model integrating complete ensemble empirical mode decomposition with adaptive noise （CEEMDAN） and the Informer deep learning model， to predict the daily average concentration of NO₂. First， CEEMDAN was employed to decompose the original NO₂ time series， and a grey wolf optimization （GWO） algorithm combined with minimum sample entropy （SE） was utilized as a fitness function to optimize the key parameters of CEEMDAN. This approach effectively reduced random fluctuations in the data and extracted intrinsic mode functions （IMFs）. Subsequently， the Pearson correlation coefficient and Spearman rank correlation coefficient were used to analyze the association strength among various pollutants， allowing for the selection of optimal features for each IMF and the construction of a high-coupling feature matrix. These features were then input into an optimized Informer model based on the Transformer architecture for encoding and modeling. By employing mechanisms such as Multi-head ProbSparse self-attention and attention distillation， the model's prediction accuracy and efficiency got enhanced. Finally， the predicted results of each IMF were summed to reconstruct the final predicted value. Experimental results indicated that the proposed model achieved an average absolute error （MAE） of 5.027 and a root mean square error （RMSE） of 6.818， demonstrating a significant improvement in prediction accuracy， compared to the original Informer model and other benchmark models including LSTM， GRU， Transformer， and SVR. The innovations in data processing， feature engineering， and model architecture presented in this study offer a more precise method for NO₂ concentration prediction， with broad application prospects that can provide robust support for regional air quality forecasting， early warning systems， and pollution control strategies.
- NO₂ concentration prediction /
- deep learning /
- Informer /
- CEEMDAN /
- feature selection

HTML全文

参考文献(48)

[1]	FINLAYSON-PITTS B J，PITTS JR J N. Chemistry of the upper and lower atmosphere：theory，experiments，and applications［M］. San Diego：Academic Press（Elsevier），1999.
[2]	GU B，ZHANG L，VAN DINGENEN R，et al. Abating ammonia is more cost-effective than nitrogen oxides for mitigating PM_{2. 5} air pollution［J］. Science，2021，374（6568）：758-762.
[3]	GAO J H，ZHU B，WANG Y Z，et al. Distribution and long-term variation of tropospheric NO₂ over China during 2005 to 2013［J］. China Environmental Science，2015，35（8）：2307-2318. 高晋徽，朱彬，王言哲，等. 2005～2013年中国地区对流层二氧化氮分布及变化趋势［J］. 中国环境科学，2015，35（8）：2307-2318.
[4]	World Health Organization. WHO global air quality guidelines：particulate matter（PM2.5 and PM₁₀），ozone，nitrogen dioxide，sulfur dioxide and carbon monoxide［M］. World Health Organization，2021.
[5]	MENG X，LIU C，CHEN R，et al. Short term associations of ambient nitrogen dioxide with daily total，cardiovascular，and respiratory mortality：multilocation analysis in 398 cities［J］. BMJ，2021，372.
[6]	FU Q R. Research on haze prediction based on multivariate linear regression［J］. Computer Science，2016，43（S6）：526-528. 付倩娆. 基于多元线性回归的雾霾预测方法研究［J］. 计算机科学，2016，43（增刊1）：526-528.
[7]	YIN Q，WANG J，HU M，et al. Estimation of daily PM_2.5 concentration and its relationship with meteorological conditions in Beijing［J］. Journal of Environmental Sciences，2016，48：161-168.
[8]	MENG F Q. Application of ARIMA model in air pollution index prediction［J］. Statistics& Decision，2009（7）：33-35. 孟凡强. ARIMA 模型在空气污染指数预测中的应用［J］. 统计与决策，2009（7）：33-35.
[9]	KUMAR A，GOYAL P. Forecasting of daily air quality index in Delhi［J］. Science of the Total Environment，2011，409（24）：5517-5523.
[10]	XIONG P，YAN W，WANG G，et al. Grey extended prediction model based on IRLS and its application on smog pollution［J］. Applied Soft Computing，2019，80：797-809.
[11]	XIONG P P，YUAN W Y，YE L L，et al. Construction of gray MGM（1，m，N）model and its prediction application on haze［J］. Systems Engineering-Theory& Practice，2020，40（3）：771-782. 熊萍萍，袁玮莹，叶琳琳，等. 灰色MGM（1，m，N）模型的构建及其在雾霾预测中的应用［J］. 系统工程理论与实践，2020，40（3）：771-782.
[12]	WU L，ZHAO H. Using FGM（1，1）model to predict the number of the lightly polluted day in Jing-Jin-Ji region of China［J］. Atmospheric Pollution Research，2019，10（2）：552-555.
[13]	NOPMONGCOL U，KOO B，TAI E，et al. Modeling Europe with CAMx for the air quality model evaluation international initiative（AQMEII）［J］. Atmospheric Environment，2012，53：177-185.
[14]	GE B Z，WANG Z F，XU X B，et al. Wet deposition of acidifying substances in different regions of China and the rest of East Asia：Modeling with updated NAQPMS［J］. Environmental Pollution，2014，187：10-21.
[15]	WANG P，QIAO X，ZHANG H. Modeling PM_{2. 5} and O₃ with aerosol feedbacks using WRF/Chem over the Sichuan Basin，southwestern China［J］. Chemosphere，2020，254：126735.
[16]	LIU C，HUANG J，HU X M，et al. Evaluation of WRF-Chem simulations on vertical profiles of PM_{2. 5} with UAV observations during a haze pollution event［J］. Atmospheric Environment，2021，252：118332.
[17]	BUONOCORE J J，DONG X，SPENGLER J D，et al. Using the Community Multiscale Air Quality（CMAQ）model to estimate public health impacts of PM_{2. 5} from individual power plants［J］. Environment international，2014，68：200-208.
[18]	LI R，MEI X，WEI L，et al. Study on the contribution of transport to PM_{2. 5} in typical regions of China using the regional air quality model RAMS-CMAQ［J］. Atmospheric environment，2019，214：116856.
[19]	LEONG W C，KELANI R O，AHMAD Z. Prediction of air pollution index（API）using support vector machine（SVM）［J］. Journal of Environmental Chemical Engineering，2020，8（3）：103208.
[20]	ZHOU Y，CHANG F J，CHANG L C，et al. Multi-output support vector machine for regional multi-step-ahead PM_{2. 5} forecasting［J］. Science of the Total Environment，2019，651：230-240.
[21]	MASMOUDI S，ELGHAZEL H，TAIEB D，et al. A machine-learning framework for predicting multiple air pollutants' concentrations via multi-target regression and feature selection［J］. Science of the Total Environment，2020，715：136991.
[22]	ZHAN Y，LUO Y，DENG X，et al. Spatiotemporal prediction of daily ambient ozone levels across China using random forest for human exposure assessment［J］. Environmental Pollution，2018，233：464-473.
[23]	ANTANASIJEVIĆ D Z，POCAJT V V，POVRENOVIĆ D S，et al. PM₁₀ emission forecasting using artificial neural networks and genetic algorithm input variable optimization［J］. Science of the Total Environment，2013，443：511-519.
[24]	LI Y，PENG T，HUA L，et al. Research and application of an evolutionary deep learning model based on improved grey wolf optimization algorithm and DBN-ELM for AQI prediction［J］. Sustainable Cities and Society，2022，87：104209.
[25]	CHAUHAN R，KAUR H，ALANKAR B. Air quality forecast using convolutional neural network for sustainable development in urban environments［J］. Sustainable Cities and Society，2021，75：103239.
[26]	MA J，CHENG J C P，LIN C，et al. Improving air quality prediction accuracy at larger temporal resolutions using deep learning and transfer learning techniques［J］. Atmospheric Environment，2019，214：116885.
[27]	MENARES C，PEREZ P，PARRAGUEZ S，et al. Forecasting PM_{2. 5} levels in Santiago de Chile using deep learning neural networks［J］. Urban Climate，2021，38：100906.
[28]	HUANG G，LI X，ZHANG B，et al. PM_{2. 5} concentration forecasting at surface monitoring sites using GRU neural network based on empirical mode decomposition［J］. Science of the Total Environment，2021，768：144516.
[29]	YU S T，LIU P. Long short-term memory-convolution neural network（LSTM-CNN）for prediction of PM_2.5 concentration in Beijing［J］. Environmental Engineering，2020，38（6）：176-180. 于伸庭，刘萍. 基于长短期记忆网络-卷积神经网络（LSTM-CNN）的北京市PM_2.5浓度预测［J］. 环境工程，2020，38（6）：176-180.
[30]	VASWANI A. Attention is all you need［J］. Advances in Neural Information Processing Systems，2017.
[31]	TAO C，JIA M，WANG G，et al. Time-sensitive prediction of NO₂ concentration in China using an ensemble machine learning model from multi-source data［J］. Journal of Environmental Sciences，2024，137：30-40.
[32]	ALSHEHHI A，WELSCH R. Artificial intelligence for improving Nitrogen Dioxide forecasting of Abu Dhabi environment agency ground-based stations［J］. Journal of Big Data，2023，10（1）：92.
[33]	ZHOU H，ZHANG S，PENG J，et al. Informer：Beyond efficient transformer for long sequence time-series forecasting［C］// Proceedings of the AAAI conference on artificial intelligence. 2021，35（12）：11106-11115.
[34]	SHANG J，ZHAO B，HUA H，et al. Application of informer model based on SPEI for drought forecasting［J］. Atmosphere，2023，14（6）：951.
[35]	HUANG X，JIANG A. Wind power generation forecast based on multi-step informer network［J］. Energies，2022，15（18）：6642.
[36]	AL-QANESS M A A，DAHOU A，EWEES A A，et al. ResInformer：residual transformer-based artificial time-series forecasting model for PM_{2. 5} concentration in three major Chinese cities［J］. Mathematics，2023，11（2）：476.
[37]	HUANG G，LI X，ZHANG B，et al. PM_{2. 5} concentration forecasting at surface monitoring sites using GRU neural network based on empirical mode decomposition［J］. Science of the Total Environment，2021，768：144516.
[38]	SONG C，FU X. Research on different weight combination in air quality forecasting models［J］. Journal of Cleaner Production，2020，261：121169.
[39]	WU Z，HUANG N E. Ensemble empirical mode decomposition：a noise-assisted data analysis method［J］. Advances in adaptive data analysis，2009，1（1）：1-41.
[40]	YEH J R，SHIEH J S，HUANG N E. Complementary ensemble empirical mode decomposition：A novel noise enhanced data analysis method［J］. Advances in adaptive data analysis，2010，2（2）：135-156.
[41]	TORRES M E，COLOMINAS M A，SCHLOTTHAUER G，et al. A complete ensemble empirical mode decomposition with adaptive noise［C］// 2011 IEEE international conference on acoustics，speech and signal processing（ICASSP）. IEEE，2011：4144-4147.
[42]	LIN Y，LIN Z，LIAO Y，et al. Forecasting the realized volatility of stock price index：A hybrid model integrating CEEMDAN and LSTM［J］. Expert Systems with Applications，2022，206：117736.
[43]	MIRJALILI S，MIRJALILI S M，LEWIS A. Grey wolf optimizer［J］. Advances in engineering software，2014，69：46-61.
[44]	RICHMAN J S，MOORMAN J R. Physiological time-series analysis using approximate entropy and sample entropy［J］. American journal of physiology-heart and circulatory physiology，2000，278（6）：H2039-H2049.
[45]	LI G，ZHANG A，ZHANG Q，et al. Pearson correlation coefficient-based performance enhancement of broad learning system for stock price prediction［J］. IEEE Transactions on Circuits and Systems II：Express Briefs，2022，69（5）：2413-2417.
[46]	JIANG J，ZHANG X，YUAN Z. Feature selection for classification with Spearman’s rank correlation coefficient-based self-information in divergence-based fuzzy rough sets［J］. Expert Systems with Applications，2024，249：123633.
[47]	BAHDANAU D. Neural machine translation by jointly learning to align and translate［J］. arXiv preprint arXiv：1409.0473，2014.
[48]	WANG X M，WANG F W，CHEN T，et al. PM_2.5 concentration prediction and uncertainty analysis based on a composite model［J］. Environmental Engineering，2020，38（8）：229-235. 王学梅，王凤文，陈滔，等. 基于组合模型的PM_2.5浓度预测及其不确定性分析［J］. 环境工程，2020，38（8）：229-235.