基于长短期记忆网络（LSTM）的多污染物时序预测模型研究

周永泉; 庄佳炜; 王川; 欧阳创; 赵春龙; 林坤森; 赵由才

doi:10.13205/j.hjgc.202604025

基于长短期记忆网络（LSTM）的多污染物时序预测模型研究

doi: 10.13205/j.hjgc.202604025

1. 上海环境卫生工程设计院有限公司, 上海 200000;
2. 福建师范大学环境与资源学院、碳中和现代产业学院, 福州 350000;
3. 同济大学环境科学与工程学院, 上海 200000

基金项目:

2025年上海市绿容局“揭榜挂帅”项目“湿垃圾厂智慧低耗除臭系统集成与应用示范”（J202502）；上海城投（集团）有限公司科技创新计划项目“基于深度学习的智能化除臭系统优化研究”（CIKY-PIRC-2024-002-001-005）

详细信息

作者简介:
周永泉(1995—),男,工程师,主要研究方向为有机固废资源化处理与恶臭污染控制。zhouyongquan@huanke.com.cn

通讯作者:
林坤森(1992—),男,副教授,主要研究方向为深度学习应用于固体废物资源化、新污染物和恶臭智能识别。kslin@fjnu.edu.cn

计量
- 文章访问数: 0
- HTML全文浏览量: 0
- PDF下载量: 0
- 被引次数: 0
出版历程
- 收稿日期: 2025-10-20
- 网络出版日期: 2026-06-06
- 刊出日期: 2026-04-01

A multi-pollutant time-series prediction model based on LSTM networks

1. Shanghai Environmental Sanitation Engineering Design Institute Co., Ltd., Shanghai 200000, China;
2. College of Environmental and Resource Sciences & College of Carbon Neutral Modern Industry, Fujian Normal University, Fuzhou 350000, China;
3. College of Environmental Science and Engineering, Tongji University, Shanghai 200000, China

摘要

摘要: 针对垃圾处理设施对臭气与多污染物的分钟级预警需求，提出了一套面向源—界（车间、厂界）场景的多变量短时序预测框架。以5 s分辨率的连续监测数据为基础，采用滑动窗口与递推多步策略的长短期记忆网络（LSTM）对恶臭浓度（OU）及VOCs、NH₃、H₂S、CH₃SH进行联合建模；建立与监管实践相衔接的评估体系，涵盖MAE、RMSE、R²、相对持久性基线的技能评分（skill score）以及阈值化误差分层，以刻画峰值段不确定性。结果显示：在工况较稳定的车间点位，VOCs、NH₃、H₂S、CH₃SH均取得较高拟合度与低误差；厂界受羽流到达时滞与扩散-稀释非平稳性的影响，OU与VOCs在峰值时段误差显著放大，技能评分在部分点位对基线优势不稳定。分层分析一致揭示非峰值段明显优于峰值段，事件驱动是主要误差来源。基于此，提出将风速风向、通风/门禁与作业节拍等外生变量以及峰值敏感损失并入模型，以提升对脉冲过程的刻画与预警能力。该研究形成了可复用的分钟级多污染物预测方法学基线与评估范式，为垃圾处理设施的运行管理与源-界联动管控提供定量支撑。
- 恶臭 /
- 长短期记忆网络 /
- 短时序列预测 /
- 技能评分 /
- 误差分层
Abstract: To meet the minute-level early-warning requirements for odor and multi-pollutant emissions at waste treatment facilities， this study proposed a multivariate short-term time-series prediction framework applicable to multi-tier scenarios covering source and boundary points （i.e.， workshops and plant boundaries）. Based on continuous online monitoring data with a 5-second resolution， a long short-term memory （LSTM） model using a sliding-window and recursive multi-step prediction strategy was constructed to jointly model odor concentration （OU） and pollutants including VOCs， NH₃， H₂S， and CH₃SH （mg/m³）. An evaluation protocol aligned with environmental supervision practice was established， incorporating mean absolute error （MAE）， root mean square error （RMSE）， goodness-of-fit （R²）， skill scores （SS） relative to a persistence baseline， and threshold-based error stratification to characterize uncertainty during peak emission periods. The results showed that at workshop monitoring sites with relatively stable operating conditions， VOCs， NH₃， H₂S， and CH₃SH exhibited a high goodness of fit and low prediction errors. In contrast， at boundary sites affected by plume arrival delays and diffusion-dilution non-stationarity， OU and VOCs displayed significantly amplified errors during peak episodes， and the skill score advantage over the baseline became unstable at certain sites. Stratified analysis consistently revealed that non-peak periods outperformed peak periods， indicating that event-driven fluctuations were the main sources of error. Accordingly， this study suggested incorporating exogenous variables such as wind speed and direction， ventilation and gate access control， and operational rhythms， along with peak-sensitive loss functions， into the model to enhance its capacity to characterize and provide early warnings for transient emission pulses. Overall， this study established a reusable methodological baseline and evaluation paradigm for minute-scale multi-pollutant prediction， providing quantitative support for the operational management and source-to-boundary coordinated control of waste treatment facilities.
- odor /
- Long Short-Term Memory （LSTM） /
- short-term time-series prediction /
- skill score /
- error stratification

HTML全文

参考文献(26)

[1]	Ministry of Ecology and Environment. Emission standards for odor pollutants:GB 14554—93［S］. Beijing：China Standards Publishing House，1993. 生态环境部. 恶臭污染物排放标准：GB 14554—93［S］. 北京：中国标准出版社，1993.
[2]	Ministry of Ecology and Environment. Ambient air quality-determination of odor-triple comparison odor bag method：GB/T 14675—93［S］. Beijing：China Standards Publishing House，1993. 生态环境部. 空气质量恶臭的测定三点比较式臭袋法：GB/T 14675—93［S］. 北京：中国标准出版社，1993.
[3]	Ministry of Ecology and Environment. Technical specifications for environmental monitoring of odor pollution：HJ 905—2017［S］. Beijing：China Environmental Publishing House，2017. 生态环境部. 恶臭污染环境监测技术规范：HJ 905—2017［S］. 北京：中国环境出版社，2017.
[4]	Ministry of Ecology and Environment. Ambient air and waste gas-Determination of odor-Triangle odor bag method：HJ 1262—2022［S］. Beijing：China Environmental Publishing House，2022. 生态环境部. 环境空气和废气臭气的测定三点比较式臭袋法：HJ 1262—2022［S］. 北京：中国环境出版社，2022.
[5]	HOCHREITER S，SCHMIDHUBER J. Long short-term memory［J］. Neural Computation，1997，9（8）：1735- 1780.
[6]	ZHANG B，RONG Y，YONG R，et al. Deep learning for air pollutant concentration prediction：a review［J］. Atmospheric Environment，2022，290：119347.
[7]	KANG J F，TAN J L，FANG L，et al. Short-term PM_2.5 concentration prediction based on XGBoost and LSTM variable weight combination model：a case study of Shanghai［J］. China Environmental Science，2021，41（9）：4016- 4025. 康俊锋，谭建林，方雷，等. XGBoost-LSTM变权组合模型支持下短期PM_2.5浓度预测：以上海为例［J］. 中国环境科学，2021，41（9）：4016- 4025.
[8]	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.
[9]	ZOU S L，REN X C，WANG C G，et al. Impacts of temporal resolution and spatial informationon neural-network-based PM_2.5 prediction model［J］. Acta Scientiarum Naturalium Universitatis Pekinensi，2020，56（3）：417- 426. 邹思琳，任晓晨，王成功，等. 时间精度与空间信息对神经网络模型预报PM_2.5浓度的影响［J］. 北京大学学报（自然科学版），2020，56（3）：417- 426.
[10]	KANG J H，SONG J，YOO S S，et al. Prediction of odor concentration emitted from wastewater treatment plant using an artificial neural network（ANN）［J］. Atmosphere，2020，11（8）：784.
[11]	LU B Q，ZHANG S Y，LIU C，et al. Machine learning-enabled estimation and high-resolution forecasting of atmospheric VOCs［J］. Atmospheric Environment，2025，358：121364.
[12]	MENDIZABAL J，VERNON D，MARTIN B，et al. Short-term memory artificial neural network modelling to predict concrete corrosion in wastewater treatment plant inlet chambers using sulphide sensors［J］. Journal of Water Process Engineering，2025，69：106821.
[13]	NARAYANAN D，BHAT M，PAUL N R，et al. Artificial intelligence driven advances in wastewater treatment：evaluating techniques for sustainability and efficacy in global facilities［J］. Desalination and Water Treatment，2024，320：100618.
[14]	LILHORE U K，SIMAIYA S，SINGH R K，et al. Advanced air quality prediction using multimodal data and dynamic modeling techniques［J］. Scientific Reports，2025，15：27867.
[15]	CZARNOTA J，MASŁOŃ A，PAJURA R. Wastewater treatment plants as a source of malodorous substances hazardous to health，including a case study from Poland［J］. International Journal of Environmental Research and Public Health，2023，20（7）：5379.
[16]	PRUDENZA S，BAX C，CAPELLI L. Implementation of an electronic nose for real-time identification of odour emission peaks at a wastewater treatment plant［J］. Heliyon，2023，9（10）：e20437.
[17]	BURGUÉS J，DOÑATE S，ESCLAPEZ M D，et al. Characterization of odour emissions in a wastewater treatment plant using a drone-based chemical sensor system［J］. Science of the Total Environment，2022，846：157290.
[18]	RATTI C，BAX C，LOTESORIERE B J，et al. Real-time monitoring of odour emissions at the fenceline of a waste treatment plant by instrumental odour monitoring systems:focus on training methods［J］. Sensors，2024，24（11）：3506.
[19]	MALINGS C，KNOWLAND K E，KELLER C A，et al. Sub-city scale hourly air quality forecasting by combining models，satellite observations，and ground measurements［J］. Earth and Space Science，2021（8）：e2021EA001743.
[20]	HYNDMAN R J，KOEHLER A B. Another look at measures of forecast accuracy［J］. International Journal of Forecasting，2006，22（4）：679- 688.
[21]	WILLMOTT C J，MATSUURA K. Advantages of the mean absolute error（MAE）over the root mean square error（RMSE）in assessing average model performance［J］. Climate Research，2005，30（1）：79- 82.
[22]	MAKRIDAKIS S，HIBON M. The M3-Competition：results，conclusions and implications［J］. International Journal of Forecasting，2000，16（4）：451- 476.
[23]	LI X，PENG L，YAO X，et al. Long short-term memory neural network for air pollutant concentration predictions:method development and evaluation［J］. Environmental Pollution，2017，231（1）：997- 1004.
[24]	CHANG Y S，CHIAO H T，ABIMANNAN S，et al. An LSTM-based aggregated model for air pollution forecasting［J］. Atmospheric Pollution Research，2020，11（8）：1451- 1463.
[25]	WU C D，LIU J M，ZHAO P，et al. Evaluation of the chemical composition and correlation between the calculated and measured odour concentration of odorous gases from a landfill in Beijing，China［J］. Atmospheric Environment，2017，164：337- 347.
[26]	BAX C，SIRONI S，CAPELLI L. How can odors be measured? An overview of methods and their applications［J］. Atmosphere，2020，11（1）：92.