基于机器学习的北京市PM2.5浓度预测模型及模拟分析
|
摘要
对北京市周边8个点多个压力高度的温度、湿度和风速数据,以及北京市PM2. 5污染数据进行了分析和归一化处理,建立了反向传播神经网络(back propagation,BP)、卷积神经网络(convolutional neural network,CNN)和长短期记忆模型(long short-term memory,LSTM)对上述气象数据和污染数据进行训练,训练结果表明:反向传播神经网络模型和卷积神经网络模型对未来1 h的PM2. 5污染等级的预测准确率较低,而长短期记忆模型的准确率较高.使用长短期记忆模型预测未来1 h的PM2. 5污染值与实际值十分接近,表明北京市的PM2. 5污染与其周边地区的气象条件关系密切.通过利用长短期记忆模型对不同压力高度的气象数据进行训练和对比,得出在利用气象数据预测污染时,仅使用近地面气象数据比使用多个高度上的气象数据更加准确.
In recent years,the air quality in China has become a matter of serious concern. Among the available indicators for evaluating air quality,PM2. 5 is one of the most important. It comprises a complex mixture of extremely small particles and liquid droplets emitted into the air,whose diameters are no more than 2. 5 μm. Environments with a high PM2. 5 index are extremely harmful to human health. Once inhaled,these particles can affect the heart and lungs and cause serious health problems. Air pollution is closely related to meteorological conditions such as wind speed,wind direction,atmospheric stability,temperature,and air humidity. With the development of various machine learning methods,deep learning models based on neural networks are increasingly applied in air pollution research. In this study,the temperature,humidity,wind velocity data at different pressure altitudes from 8 locations around Beijing and average of PM2. 5 data in Beijing were analyzed and normalized. Multi-dimensional data was ideal for research applications using machine learning methods. and three neural network models were built,including the back propagation( BP),convolutional neural network( CNN),and long short-term memory( LSTM) models,and trained them using the meteorological and PM2. 5 data.The results indicate that the accuracies of the back propagation and convolutional neural network models in predicting the PM2. 5 pollution level in the next hour is much lower than that of the long short-term memory model. The PM2. 5 pollution index predicted for the next hour by the long short-term memory model is very close to the actual value. This result reveals the strong relationship between the PM2. 5 pollution index of Beijing and the local meteorological conditions. The long short-term memory model is trained using meteorological data from different pressure altitudes,and found it to be more accurate in predicting pollution levels when using near-surface meteorological data than that obtained from multiple altitudes.
In recent years,the air quality in China has become a matter of serious concern. Among the available indicators for evaluating air quality,PM2. 5 is one of the most important. It comprises a complex mixture of extremely small particles and liquid droplets emitted into the air,whose diameters are no more than 2. 5 μm. Environments with a high PM2. 5 index are extremely harmful to human health. Once inhaled,these particles can affect the heart and lungs and cause serious health problems. Air pollution is closely related to meteorological conditions such as wind speed,wind direction,atmospheric stability,temperature,and air humidity. With the development of various machine learning methods,deep learning models based on neural networks are increasingly applied in air pollution research. In this study,the temperature,humidity,wind velocity data at different pressure altitudes from 8 locations around Beijing and average of PM2. 5 data in Beijing were analyzed and normalized. Multi-dimensional data was ideal for research applications using machine learning methods. and three neural network models were built,including the back propagation( BP),convolutional neural network( CNN),and long short-term memory( LSTM) models,and trained them using the meteorological and PM2. 5 data.The results indicate that the accuracies of the back propagation and convolutional neural network models in predicting the PM2. 5 pollution level in the next hour is much lower than that of the long short-term memory model. The PM2. 5 pollution index predicted for the next hour by the long short-term memory model is very close to the actual value. This result reveals the strong relationship between the PM2. 5 pollution index of Beijing and the local meteorological conditions. The long short-term memory model is trained using meteorological data from different pressure altitudes,and found it to be more accurate in predicting pollution levels when using near-surface meteorological data than that obtained from multiple altitudes.
引文
[1] Tai A P K,Mickley L J,Jacob D J. Correlations between fine particulate matter(PM2. 5)and meteorological variables in the United States:implications for the sensitivity of PM2. 5to climate change.Atmos Environ,2010,44(32):3976
[2] Pateraki S,Asimakopoulos D N,Flocas H A,et al. The role of meteorology on different sized aerosol fractions(PM10,PM2. 5,PM2. 5-10). Sci Total Environ,2012,419:124
[3] Zhao X J,Zhang X L,Xu X F,et al. Seasonal and diurnal variations of ambient PM2. 5concentration in urban and rural environments in Beijing. Atmos Environ,2009,43(18):2893
[4] Pan B F,Zhao Y L,Li J J,et al. Analysis of the scavenging efficiency on PM2. 5concentration of some kinds of meteorological factors. Environ Sci Technol,2012,25(6):41(潘本锋,赵熠琳,李健军,等.气象因素对大气中PM2. 5的去除效应分析.环境科技,2012,25(6):41)
[5] Yan W L,Zhou D P,Wang Y F,et al. Concentrations and size distributions of inhalable particles in summer and winter in Shenyang. J Appl Meteorol Sci,2008,19(4):435(严文莲,周德平,王扬峰,等.沈阳冬夏季可吸入颗粒物浓度及尺度谱分布特征.应用气象学报,2008,19(4):435)
[6] Liu J,Zhang X L,Xu X F,et al. Comparison analysis of variation characteristics of SO2,NOx,O3and PM2. 5between rural and urban areas,Beijing. Environ Sci,2008,29(4):1059(刘洁,张小玲,徐晓峰,等.北京地区SO2、NOx、O3和PM2. 5变化特征的城郊对比分析.环境科学,2008,29(4):1059)
[7] Dong X L,Liu D M,Yuan Y S,et al. Pollution characteristics and influencing factors of atmospheric particulates in Beijing during the summer of 2005. Chin J Environ Eng,2007,1(9):100(董雪玲,刘大锰,袁杨森,等.北京市2005年夏季大气颗粒物污染特征及影响因素.环境工程学报,2007,1(9):100)
[8] Sillman S. The relation between ozone,NOxand hydrocarbons in urban and polluted rural environments. Atmos Environ,1999,33(12):1821
[9] Zhou J X. Correlative analysis on relationship between changes of several main contaminations and some meteorological elements. J Appl Meteorol Sci,2005,16(Suppl):123(周江兴.北京市几种主要污染物浓度与气象要素的相关分析.应用气象学报,2005,16(增刊):123)
[10] Zhou L,Xu X D,Ding G A,et al. The correlation factors and pollution forecast model for PM2. 5 concentration in Beijing Area.Acta Meteorol Sin,2003,61(6):761(周丽,徐祥德,丁国安,等.北京地区气溶胶PM2. 5粒子浓度的相关因子及其估算模型.气象学报,2003,61(6):761)
[11] Ong B T,Sugiura K,Zettsu K. Dynamically pre-trained deep recurrent neural networks using environmental monitoring data for predicting PM2. 5. Neural Comput Appl,2016,27(6):1553
[12] Lary D J,Faruque F S,Malakar N,et al. Estimating the global abundance of ground level presence of particulate matter(PM2. 5). Geospatial Health,2014,8(3):611
[13] Ong B T,Sugiura K,Zettsu K. Dynamic pre-training of deep recurrent neural networks for predicting environmental monitoring data∥2014 IEEE International Conference on Big Data(Big Data). Washington,DC,2014:760
[14] Zheng Y,Zhu C Z. A prediction method of atmospheric PM2. 5based on DBNs. J Shandong Univ Eng Sci,2014,44(6):19(郑毅,朱成璋.基于深度信念网络的PM2. 5预测.山东大学学报(工学版),2014,44(6):19)
[15] Dai L J,Zhang C J,Ma L M. Dynamic forecasting model of shortterm PM2. 5 concentration based on machine learning. J Comput Appl,2017,37(11):3057(戴李杰,张长江,马雷鸣.基于机器学习的PM2. 5短期浓度动态预报模型.计算机应用,2017,37(11):3057)
[16] Li X,Peng L,Hu Y,et al. Deep learning architecture for air quality predictions. Environ Sci Pollut Res,2016,23(22):22408
[17] Rumelhart D E,Mc Clelland J L. Parallel Distributed Processing:Explorations in the Microstructure of Cognition. Volume 1. Foundations. Cambridge:MIT Press,1986
[18] Hubel D H,Wiesel T N. Receptive fields of single neurones in the cat's striate cortex. J Physiol,1959,148(3):574
[19] Pigou L,Van Den Oord A,Dieleman S,et al. Beyond temporal pooling:Recurrence and temporal convolutions for gesture recognition in video. Int J Comput Vision,2018,126(2-4):430
[20] Yang J B,Nguyen M N,San P P,et al. Deep convolutional neural networks on multichannel time series for human activity recognition//Proceedings of the Twenty-Fourth International Conference on Artificial Intelligence. Buenos Aires,2015:3995
[21] Hochreiter S,Schmidhuber J. Long short-term memory. Neural Comput,1997,9(8):1735
[22] Jia M W,Kang N,Zhao T L. Characteristics oftypical autumn and winter haze pollution episodes and their boundary layer in Nanjing. Environ Sci Technol,2014,37(120):105(贾梦唯,康娜,赵天良.南京秋冬季典型霾污染过程及边界层特征分析.环境科学与技术,2014,37(120):105)
[23] Zhang J H,Xu S R,Zhao S M. Analysis of effect of temperature inversion to atmosphere pollution. Heilongjiang Environ J,2008,32(2):24(张继红,徐盛荣,赵淑敏.逆温天气对大气污染的影响分析.黑龙江环境通报,2008,32(2):24)
[24] Ning H W. An Analysis on Meteorological Factors and Study on Predicting Air Pollution in Xi'an City[Dissertation]. Nanjing:Nanjing University of Information Science and Technology,2006(宁海文.西安市大气污染气象条件分析及空气质量预报方法研究[学位论文].南京:南京信息工程大学,2006)
[25] Xiao T Y,Ren H Y,Zhang X W. Analysis and prediction of the winter inversion in Harbin. J Northeast Agric Univ,2001,32(2):139(肖同玉,任红玉,张兴文.哈尔滨冬季逆温规律分析及预报.东北农业大学学报,2001,32(2):139)
[26] Cao H L,Chen Q. Inverse temperature characteristics of boundary layer and its relationship with air pollution in Xi'an. J Shaanxi Meteorol,2014(2):13(曹红丽,陈奇.西安边界层逆温特征及其与空气污染的关系.陕西气象,2014(2):13)
[2] Pateraki S,Asimakopoulos D N,Flocas H A,et al. The role of meteorology on different sized aerosol fractions(PM10,PM2. 5,PM2. 5-10). Sci Total Environ,2012,419:124
[3] Zhao X J,Zhang X L,Xu X F,et al. Seasonal and diurnal variations of ambient PM2. 5concentration in urban and rural environments in Beijing. Atmos Environ,2009,43(18):2893
[4] Pan B F,Zhao Y L,Li J J,et al. Analysis of the scavenging efficiency on PM2. 5concentration of some kinds of meteorological factors. Environ Sci Technol,2012,25(6):41(潘本锋,赵熠琳,李健军,等.气象因素对大气中PM2. 5的去除效应分析.环境科技,2012,25(6):41)
[5] Yan W L,Zhou D P,Wang Y F,et al. Concentrations and size distributions of inhalable particles in summer and winter in Shenyang. J Appl Meteorol Sci,2008,19(4):435(严文莲,周德平,王扬峰,等.沈阳冬夏季可吸入颗粒物浓度及尺度谱分布特征.应用气象学报,2008,19(4):435)
[6] Liu J,Zhang X L,Xu X F,et al. Comparison analysis of variation characteristics of SO2,NOx,O3and PM2. 5between rural and urban areas,Beijing. Environ Sci,2008,29(4):1059(刘洁,张小玲,徐晓峰,等.北京地区SO2、NOx、O3和PM2. 5变化特征的城郊对比分析.环境科学,2008,29(4):1059)
[7] Dong X L,Liu D M,Yuan Y S,et al. Pollution characteristics and influencing factors of atmospheric particulates in Beijing during the summer of 2005. Chin J Environ Eng,2007,1(9):100(董雪玲,刘大锰,袁杨森,等.北京市2005年夏季大气颗粒物污染特征及影响因素.环境工程学报,2007,1(9):100)
[8] Sillman S. The relation between ozone,NOxand hydrocarbons in urban and polluted rural environments. Atmos Environ,1999,33(12):1821
[9] Zhou J X. Correlative analysis on relationship between changes of several main contaminations and some meteorological elements. J Appl Meteorol Sci,2005,16(Suppl):123(周江兴.北京市几种主要污染物浓度与气象要素的相关分析.应用气象学报,2005,16(增刊):123)
[10] Zhou L,Xu X D,Ding G A,et al. The correlation factors and pollution forecast model for PM2. 5 concentration in Beijing Area.Acta Meteorol Sin,2003,61(6):761(周丽,徐祥德,丁国安,等.北京地区气溶胶PM2. 5粒子浓度的相关因子及其估算模型.气象学报,2003,61(6):761)
[11] Ong B T,Sugiura K,Zettsu K. Dynamically pre-trained deep recurrent neural networks using environmental monitoring data for predicting PM2. 5. Neural Comput Appl,2016,27(6):1553
[12] Lary D J,Faruque F S,Malakar N,et al. Estimating the global abundance of ground level presence of particulate matter(PM2. 5). Geospatial Health,2014,8(3):611
[13] Ong B T,Sugiura K,Zettsu K. Dynamic pre-training of deep recurrent neural networks for predicting environmental monitoring data∥2014 IEEE International Conference on Big Data(Big Data). Washington,DC,2014:760
[14] Zheng Y,Zhu C Z. A prediction method of atmospheric PM2. 5based on DBNs. J Shandong Univ Eng Sci,2014,44(6):19(郑毅,朱成璋.基于深度信念网络的PM2. 5预测.山东大学学报(工学版),2014,44(6):19)
[15] Dai L J,Zhang C J,Ma L M. Dynamic forecasting model of shortterm PM2. 5 concentration based on machine learning. J Comput Appl,2017,37(11):3057(戴李杰,张长江,马雷鸣.基于机器学习的PM2. 5短期浓度动态预报模型.计算机应用,2017,37(11):3057)
[16] Li X,Peng L,Hu Y,et al. Deep learning architecture for air quality predictions. Environ Sci Pollut Res,2016,23(22):22408
[17] Rumelhart D E,Mc Clelland J L. Parallel Distributed Processing:Explorations in the Microstructure of Cognition. Volume 1. Foundations. Cambridge:MIT Press,1986
[18] Hubel D H,Wiesel T N. Receptive fields of single neurones in the cat's striate cortex. J Physiol,1959,148(3):574
[19] Pigou L,Van Den Oord A,Dieleman S,et al. Beyond temporal pooling:Recurrence and temporal convolutions for gesture recognition in video. Int J Comput Vision,2018,126(2-4):430
[20] Yang J B,Nguyen M N,San P P,et al. Deep convolutional neural networks on multichannel time series for human activity recognition//Proceedings of the Twenty-Fourth International Conference on Artificial Intelligence. Buenos Aires,2015:3995
[21] Hochreiter S,Schmidhuber J. Long short-term memory. Neural Comput,1997,9(8):1735
[22] Jia M W,Kang N,Zhao T L. Characteristics oftypical autumn and winter haze pollution episodes and their boundary layer in Nanjing. Environ Sci Technol,2014,37(120):105(贾梦唯,康娜,赵天良.南京秋冬季典型霾污染过程及边界层特征分析.环境科学与技术,2014,37(120):105)
[23] Zhang J H,Xu S R,Zhao S M. Analysis of effect of temperature inversion to atmosphere pollution. Heilongjiang Environ J,2008,32(2):24(张继红,徐盛荣,赵淑敏.逆温天气对大气污染的影响分析.黑龙江环境通报,2008,32(2):24)
[24] Ning H W. An Analysis on Meteorological Factors and Study on Predicting Air Pollution in Xi'an City[Dissertation]. Nanjing:Nanjing University of Information Science and Technology,2006(宁海文.西安市大气污染气象条件分析及空气质量预报方法研究[学位论文].南京:南京信息工程大学,2006)
[25] Xiao T Y,Ren H Y,Zhang X W. Analysis and prediction of the winter inversion in Harbin. J Northeast Agric Univ,2001,32(2):139(肖同玉,任红玉,张兴文.哈尔滨冬季逆温规律分析及预报.东北农业大学学报,2001,32(2):139)
[26] Cao H L,Chen Q. Inverse temperature characteristics of boundary layer and its relationship with air pollution in Xi'an. J Shaanxi Meteorol,2014(2):13(曹红丽,陈奇.西安边界层逆温特征及其与空气污染的关系.陕西气象,2014(2):13)