PREDICTION OF PM2.5 CONCENTRATION BASED ON INFORMER
-
摘要: 针对现有PM2.5浓度时序预测模型预测精度不高的问题,基于Informer建立了1个Seq2Seq的单站点PM2.5浓度多步时序预测模型,以历史污染物数据和气象数据为输入,实现对未来一段时间PM2.5浓度的预测。所构建模型基于ProbSparse (概率稀疏)自注意力机制提取所输入的序列信息,能够广泛地捕获输入序列的长期依赖信息,并对影响因子之间复杂的非线性关系进行建模,从而提高预测准确度。采用北京市2015-2019年逐小时空气污染物数据与气象数据进行模型训练、验证和测试,建立与循环神经网络(RNN)、长短期记忆网络(LSTM)的对比实验并与其他现有研究方法进行比较,结果表明:对未来1~6 h的PM2.5浓度时序预测,Informer的平均绝对误差(MAE)、均方根误差(RMSE)和可决系数(R2)指标均为最好,实现了较为准确的预测。Abstract: For improving the low accuracy of the existing models for time series prediction of PM2.5 concentration,a Seq2Seq multi-step PM2.5 concentration prediction model for single-site based on Informer was proposed.With a series of air pollutant data and meteorological data in the past,Informer could make a forecast for PM2.5 concentration in the future.The constructed model extracted the information of the input sequence based on the probsparse self-attention mechanism,which could widely capture the long-range dependency of the input sequence and model the complex nonlinearity between features,to improve the prediction accuracy eventually.The hourly air pollutant data and meteorological data of Beijing from 2015 to 2019 were used for training,validation and testing.Compared with RNN,LSTM and other existing models,the MAE,RMSE and R2 metrics of Informer were the best for the time series prediction of PM2.5 concentration in the next 1 to 6 hours,and then a more accurate prediction was realized.
-
Key words:
- PM2.5 concentration prediction /
- machine learning /
- Informer /
- self-attention mechanism /
- time series
-
[1] 郭新彪,魏红英.大气PM2.5对健康影响的研究进展[J].科学通报,2013,58(13):1171-1177. [2] 宋宇,唐孝炎,方晨,等.北京市能见度下降与颗粒物污染的关系[J].环境科学学报,2003,23(4):468-471. [3] 师华定,高庆先,张时煌,等.空气污染对气候变化影响与反馈的研究评述[J].环境科学研究,2012,25(9):974-980. [4] 蒋锋,乔雅倩.基于样本熵和优化极限学习机的PM2.5浓度预测[J].统计与决策,2021,37(3):166-171. [5] 秦思达,王帆,王堃,等.基于WRF-CMAQ模型的辽宁中部城市群PM2.5化学组分特征[J].环境科学研究,2021,34(6):1277-1286. [6] 杜勃莹,马云峰,王琦,等.基于WRF-Chem模型的沈阳市颗粒物扩散特征和成因分析[J].环境工程,2021,39(2):89-97,104. [7] 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. [8] 刘宗伟,周彩丽,马冬梅,等.自回归移动平均模型在预测PM2.5中的应用[J].预防医学论坛,2016,22(8):582-584. [9] 朱晏民,徐爱兰,孙强.基于深度学习的空气质量预报方法新进展[J].中国环境监测,2020,36(3):10-18. [10] 李建新,刘小生,刘静,等.基于MRMR-HK-SVM模型的PM2.5浓度预测[J].中国环境科学,2019,39(6):2304-2310. [11] 段大高,赵振东,梁少虎,等.基于LSTM的PM2.5浓度预测模型[J].计算机测量与控制,2019,27(3):215-219. [12] 赵文芳,林润生,唐伟,等.基于深度学习的PM2.5短期预测模型[J].南京师大学报(自然科学版),2019,42(3):32-41. [13] SCHUSTER M, PALIWAL K K. Bidirectional recurrent neural networks[J]. IEEE transactions on Signal Processing, 1997, 45(11):2673-2681. [14] HOCHREITER S, SCHMIDHUBER J. Long short-term memory[J]. Neural Computation, 1997, 9(8):1735-1780. [15] ZHOU H, ZHANG S, PENG J, et al. Informer:beyond efficient transformer for long sequence time-series forecasting[C]//Proceedings of AAAI, 2021. [16] VASWANI A, SHAZEER N, PARMAR N, et al. Attention is all you need[C]//Advances in Neural Information Processing Systems, 2017:5998-6008. [17] 陆瑶,杨洁,邵智娟,等.基于阶段式时序注意力网络的PM2.5鲁棒预测[J].环境工程,2021,39(10):93-100. [18] 张婷婷.北京典型地区大气PM2.5的理化特征及来源初步解析[D].北京:北京建筑大学,2018. [19] 于建华,虞统,魏强,等.北京地区PM10和PM2.5质量浓度的变化特征[J].环境科学研究,2004,17(1):45-47. [20] XIE Y Y,ZHAO B,ZHANG L, et al.Spatiotemporal variations of PM2.5 and PM10 concentrations between 31 Chinese cities and their relationships with SO2,NO2,CO and O3[J].Particuology,2015,20(3):141-149. [21] 杨兴川,赵文吉,熊秋林,等.2016年京津冀地区PM2.5时空分布特征及其与气象因素的关系[J].生态环境学报,2017,26(10):1747-1754. [22] 黄春桃,范东平,卢集富,等.基于深度学习的广州市大气PM2.5和PM10浓度预测[J].环境工程,2021,39(12):135-140. [23] 王菲. LSTM循环神经网络的研究进展与应用[D].哈尔滨:黑龙江大学,2021. [24] 张宸鹏.回复式神经网络若干关键问题研究[D].成都:电子科技大学,2021. [25] 赵滨,刘斌.基于Stacking的地面PM2.5浓度估算[J].环境工程,2020,38(2):153-159. [26] 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PM2.5浓度预测[J].环境工程,2020,38(6):176-180,66. [27] 李晓理,梅建想,张山.基于改进粒子群优化BP_Adaboost神经网络的PM2.5浓度预测[J].大连理工大学学报,2018,58(3):316-323.
点击查看大图
计量
- 文章访问数: 513
- HTML全文浏览量: 94
- PDF下载量: 30
- 被引次数: 0