支持向量机回归在臭氧预报中的应用
|
摘要
采用南京工业区2016年5月20日～8月15日这一高臭氧(O_3)期的O_3、O_3前体物和常规气象资料数据,利用支持向量机回归(SVMr)方法分别预报O_3的小时值、日最大值和最大8 h滑动平均值.结果表明,O_3小时值预报的相关系数(R~2)为0. 84,平均绝对误差(MAE)和平均绝对百分误差(MAPE)分别为3. 44×10-9和24. 48,O_3前期浓度、紫外B波段辐射(UVB)和NO_2浓度是关键因子. O_3日最大值预报的主要因子是NO_x在07:00的浓度和UVB.预报O_38 h时UVB和气温起重要作用.加入前体物项能够使O_3的预报精度提升10%～28%.与多元线性回归方法相比,SVMr对O_3浓度的预报有明显优势.
Support vector machine regression( SVMr) was proposed to forecast hourly ozone( O_3) concentrations,daily maximum O_3 concentrations,and maximum 8 h moving average O_3 concentrations( O_38 h) by employing the observations of meteorological variables and O_3 and its precursors during the high O_3 periods from May 20 to August 15,2016 at an industrial area in Nanjing. The squared correlation coefficient( R~2) of the hourly O_3 concentrations forecast was 0. 84. The mean absolute error( MAE) and mean absolute percentage error( MAPE) were 3. 44 × 10-9 and 24. 48,respectively. The key factors for the hourly O_3 forecast were the O_3 preconcentrations,amount of ultraviolet radiation B( UVB),and the NO_2 concentration. The main factors for the O_3 daily maximum forecast were the NO_xconcentrations at 07: 00 and the UVB level. Temperature and UVB played an important role in predicting O_38 h.In general,taking precursors into account could increase the accuracy of O_3 prediction by 10%-28%. For O_3 concentration forecasting,SVMr gave significantly better predictions than multiple linear regression methods.
Support vector machine regression( SVMr) was proposed to forecast hourly ozone( O_3) concentrations,daily maximum O_3 concentrations,and maximum 8 h moving average O_3 concentrations( O_38 h) by employing the observations of meteorological variables and O_3 and its precursors during the high O_3 periods from May 20 to August 15,2016 at an industrial area in Nanjing. The squared correlation coefficient( R~2) of the hourly O_3 concentrations forecast was 0. 84. The mean absolute error( MAE) and mean absolute percentage error( MAPE) were 3. 44 × 10-9 and 24. 48,respectively. The key factors for the hourly O_3 forecast were the O_3 preconcentrations,amount of ultraviolet radiation B( UVB),and the NO_2 concentration. The main factors for the O_3 daily maximum forecast were the NO_xconcentrations at 07: 00 and the UVB level. Temperature and UVB played an important role in predicting O_38 h.In general,taking precursors into account could increase the accuracy of O_3 prediction by 10%-28%. For O_3 concentration forecasting,SVMr gave significantly better predictions than multiple linear regression methods.
引文
[1] Karlsson P E,Klingberg J,Engardt M,et al. Past,present and future concentrations of ground-level ozone and potential impacts on ecosystems and human health in northern Europe[J]. Science of the Total Environment,2017,576:22-35.
[2] Wang T,Xue L K,Brimblecombe P,et al. Ozone pollution in China:a review of concentrations, meteorological influences,chemical precursors, and effects[J]. Science of the Total Environment,2016,575:1582-1596.
[3] Yang C X,Yang H B,Guo S,et al. Alternative ozone metrics and daily mortality in Suzhou:the China air pollution and health effects study(CAPES)[J]. Science of the Total Environment,2012,426:83-89.
[4] Zhai L, Li S, Zou B, et al. An improved geographically weighted regression model for PM2. 5concentration estimation in large areas[J]. Atmospheric Environment,2018,181:145-154.
[5]安俊琳,王跃思,朱彬.主成分和回归分析方法在大气臭氧预报的应用——以北京夏季为例[J].环境科学学报,2010,30(6):1286-1294.An J L,Wang Y S,Zhu B. Principal component and multiple regression analysis predicting ozone concentrations:case study in summer in Beijing[J]. Acta Scientiae Circumstantiae,2010,30(6):1286-1294.
[6] Barrero M A,Grimalt J O,Cantón L. Prediction of daily ozone concentration maxima in the urban atmosphere[J].Chemometrics and Intelligent Laboratory Systems,2006,80(1):67-76.
[7] Gao M,Yin L T,Ning J C. Artificial neural network model for ozone concentration estimation and Monte Carlo analysis[J].Atmospheric Environment,2018,184:129-139.
[8] Mao X,Shen T,Feng X. Prediction of hourly ground-level PM2. 5concentrations 3 days in advance using neural networks with satellite data in eastern China[J]. Atmospheric Pollution Research,2017,8(6):1005-1015.
[9] Gómez-Sanchis J,Martín-Guerrero J D,Soria-Olivas E,et al.Neural networks for analysing the relevance of input variables in the prediction of tropospheric ozone concentration[J].Atmospheric Environment,2006,40(32):6173-6180.
[10]丁愫,陈报章,王瑾,等.基于决策树的统计预报模型在臭氧浓度时空分布预测中的应用研究[J].环境科学学报,2018,38(8):3229-3242.Ding S,Chen B Z,Wang J,et al. An applied research of decision-tree based statistical model in forecasting the spatialtemporal distribution of O3[J]. Acta Scientiae Circumstantiae,2018,38(8):3229-3242.
[11]靳小兵,徐会明,卜俊伟,等.决策树方法在一次历史异常雷电活动中的预报能力检验[J].高原山地气象研究,2013,33(4):56-60,78.Jin X B,Xu H M,Pu J W,et al. The forecast ability test of decision tree method in an abnormal lightning activity[J].Plateau and Mountain Meteorology Research,2013,33(4):56-60,78.
[12] Vapnik V N. The nature of statistical learning theory[M]. New York:Springer,2000. 17-34.
[13] García Nieto P J,Combarro E F,del Coz Díaz J J,et al. A SVM-based regression model to study the air quality at local scale in Oviedo urban area(Northern Spain):a case study[J].Applied Mathematics and Computation,2013,219(17):8923-8937.
[14] Lu W Z,Wang W J. Potential assessment of the“support vector machine”method in forecasting ambient air pollutant trends[J].Chemosphere,2005,59(5):693-701.
[15] Quej V H,Almorox J,Arnaldo J A,et al. ANFIS,SVM and ANN soft-computing techniques to estimate daily global solar radiation in a warm sub-humid environment[J]. Journal of Atmospheric and Solar-Terrestrial Physics,2017,155:62-70.
[16]赵文婧,赵中军,汪结华,等.基于支持向量机的云量精细化预报研究[J].干旱气象,2016,34(3):568-574,589.Zhao W J,Zhao Z J,Wang J H,et al. A study on refined forecast of cloud cover based on support vector machine[J].Journal of Arid Meteorology,2016,34(3):568-574,589.
[17]熊秋芬,顾永刚,王丽.支持向量机分类方法在天空云量预报中的应用[J].气象,2007,33(5):20-26.Xiong Q F,Gu Y G,Wang L. Application of SVM method to cloud amount forecast[J]. Meteorological Monthly,2007,33(5):20-26.
[18]吴波,胡邦辉,王学忠,等.基于近似支持向量机的能见度释用预报研究[J].热带气象学报,2017,33(1):104-110.Wu B,Hu B H,Wang X Z,et al. Visibility forecast based on proximal support vector machine[J]. Journal of Tropical Meteorology,2017,33(1):104-110.
[19]郑朝霞,周梅,季致建,等. SVM方法在霾识别和能见度预报中的应用[J].气象科技进展,2016,6(6):30-34.Zheng Z X,Zhou M,Ji Z J,et al. Application of SVM method to identification of haze and prediction of visibility[J]. Advances in Meteorological Science and Technology,2016,6(6):30-34.
[20] Canu S,Rakotomamonjy A. Ozone peak and pollution forecasting using support vectors[R]. Yokohama:IFAC,2001.
[21] García Nieto P J,García-Gonzalo E,Sánchez A B,et al. Air quality modeling using the PSO-SVM-based approach, MLP neural network,and M5 model tree in the metropolitan area of Oviedo(Northern Spain)[J]. Environmental Modeling&Assessment,2018,23(3):229-247.
[22] García Nieto P J,Sánchez Lasheras F,García-Gonzalo E,et al.PM10concentration forecasting in the metropolitan area of Oviedo(Northern Spain)using models based on SVM,MLP,VARMA and ARIMA:a case study[J]. Science of the Total Environment,2018,621:753-761.
[23] Yeganeh B,Motlagh M S P,Rashidi Y,et al. Prediction of CO concentrations based on a hybrid partial least square and support vector machine model[J]. Atmospheric Environment,2012,55:357-365.
[24]朱佳,王振会,金天力,等.基于小波分解和最小二乘支持向量机的大气臭氧含量时间序列预测[J].气候与环境研究,2010,15(3):295-302.Zhu J,Wang Z H,Jin T L,et al. Combination of wavelet decomposition and least square support vector machine to forcast atmospheric ozone content time series[J]. Climatic and Environmental Research,2010,15(3):295-302.
[25] Niu M F,Gan K,Sun S L,et al. Application of decompositionensemble learning paradigm with phase space reconstruction for day-ahead PM2. 5concentration forecasting[J]. Journal of Environmental Management,2017,196:110-118.
[26] Feng Y,Zhang W F,Sun D Z,et al. Ozone concentration forecast method based on genetic algorithm optimized back propagation neural networks and support vector machine data classification[J]. Atmospheric Environment,2011,45(11):1979-1985.
[27] Xu Y Z,Yang W D,Wang J Z. Air quality early-warning system for cities in China[J]. Atmospheric Environment,2017,148:239-257.
[28]王俊秀,安俊琳,邵平,等.南京北郊大气臭氧周末效应特征分析[J].环境科学,2017,38(6):2256-2263.Wang J X,An J L,Shao P,et al. Characteristic study on the“weekend effect” of atmospheric O3in northern Suburb of Nanjing[J]. Environmental Science,2017,38(6):2256-2263.
[29] An J L,Zhu B,Wang H L,et al. Characteristics and source apportionment of VOCs measured in an industrial area of Nanjing, Yangtze River Delta, China[J]. Atmospheric Environment,2014,97:206-214.
[30] Luna A S,Paredes M L L,de Oliveira G C G,et al. Prediction of ozone concentration in tropospheric levels using artificial neural networks and support vector machine at Rio de Janeiro,Brazil[J]. Atmospheric Environment,2014,98:98-104.
[31]安俊琳,朱彬,李用宇.南京北郊大气VOCs体积分数变化特征[J].环境科学,2013,34(12):4504-4512.An J L,Zhu B,Li Y Y. Variation characteristics of ambient volatile organic compounds(VOCs)in Nanjing northern suburb,China[J]. Environmental Science,2013,34(12):4504-4512.
[32] Chelani A B. Prediction of daily maximum ground ozone concentration using support vector machine[J]. Environmental Monitoring and Assessment,2010,162(1-4):169-176.
[33]张玉欣,安俊琳,王俊秀,等.南京工业区挥发性有机物来源解析及其对臭氧贡献评估[J].环境科学,2018,39(2):502-510.Zhang Y X,An J L,Wang J X,et al. Source analysis of volatile organic compounds in the Nanjing industrial area and evaluation of their contribution to ozone[J]. Environmental Science,2018,39(2):502-510.
[34]邵平,安俊琳,杨辉,等.南京北郊夏季近地层臭氧及其前体物体积分数变化特征[J].环境科学,2014,35(11):4031-4043.Shao P,An J L,Yang H,et al. Variation characteristics of surface ozone and its precursors during summertime in Nanjing northern suburb[J]. Environmental Science,2014,35(11):4031-4043.
[2] Wang T,Xue L K,Brimblecombe P,et al. Ozone pollution in China:a review of concentrations, meteorological influences,chemical precursors, and effects[J]. Science of the Total Environment,2016,575:1582-1596.
[3] Yang C X,Yang H B,Guo S,et al. Alternative ozone metrics and daily mortality in Suzhou:the China air pollution and health effects study(CAPES)[J]. Science of the Total Environment,2012,426:83-89.
[4] Zhai L, Li S, Zou B, et al. An improved geographically weighted regression model for PM2. 5concentration estimation in large areas[J]. Atmospheric Environment,2018,181:145-154.
[5]安俊琳,王跃思,朱彬.主成分和回归分析方法在大气臭氧预报的应用——以北京夏季为例[J].环境科学学报,2010,30(6):1286-1294.An J L,Wang Y S,Zhu B. Principal component and multiple regression analysis predicting ozone concentrations:case study in summer in Beijing[J]. Acta Scientiae Circumstantiae,2010,30(6):1286-1294.
[6] Barrero M A,Grimalt J O,Cantón L. Prediction of daily ozone concentration maxima in the urban atmosphere[J].Chemometrics and Intelligent Laboratory Systems,2006,80(1):67-76.
[7] Gao M,Yin L T,Ning J C. Artificial neural network model for ozone concentration estimation and Monte Carlo analysis[J].Atmospheric Environment,2018,184:129-139.
[8] Mao X,Shen T,Feng X. Prediction of hourly ground-level PM2. 5concentrations 3 days in advance using neural networks with satellite data in eastern China[J]. Atmospheric Pollution Research,2017,8(6):1005-1015.
[9] Gómez-Sanchis J,Martín-Guerrero J D,Soria-Olivas E,et al.Neural networks for analysing the relevance of input variables in the prediction of tropospheric ozone concentration[J].Atmospheric Environment,2006,40(32):6173-6180.
[10]丁愫,陈报章,王瑾,等.基于决策树的统计预报模型在臭氧浓度时空分布预测中的应用研究[J].环境科学学报,2018,38(8):3229-3242.Ding S,Chen B Z,Wang J,et al. An applied research of decision-tree based statistical model in forecasting the spatialtemporal distribution of O3[J]. Acta Scientiae Circumstantiae,2018,38(8):3229-3242.
[11]靳小兵,徐会明,卜俊伟,等.决策树方法在一次历史异常雷电活动中的预报能力检验[J].高原山地气象研究,2013,33(4):56-60,78.Jin X B,Xu H M,Pu J W,et al. The forecast ability test of decision tree method in an abnormal lightning activity[J].Plateau and Mountain Meteorology Research,2013,33(4):56-60,78.
[12] Vapnik V N. The nature of statistical learning theory[M]. New York:Springer,2000. 17-34.
[13] García Nieto P J,Combarro E F,del Coz Díaz J J,et al. A SVM-based regression model to study the air quality at local scale in Oviedo urban area(Northern Spain):a case study[J].Applied Mathematics and Computation,2013,219(17):8923-8937.
[14] Lu W Z,Wang W J. Potential assessment of the“support vector machine”method in forecasting ambient air pollutant trends[J].Chemosphere,2005,59(5):693-701.
[15] Quej V H,Almorox J,Arnaldo J A,et al. ANFIS,SVM and ANN soft-computing techniques to estimate daily global solar radiation in a warm sub-humid environment[J]. Journal of Atmospheric and Solar-Terrestrial Physics,2017,155:62-70.
[16]赵文婧,赵中军,汪结华,等.基于支持向量机的云量精细化预报研究[J].干旱气象,2016,34(3):568-574,589.Zhao W J,Zhao Z J,Wang J H,et al. A study on refined forecast of cloud cover based on support vector machine[J].Journal of Arid Meteorology,2016,34(3):568-574,589.
[17]熊秋芬,顾永刚,王丽.支持向量机分类方法在天空云量预报中的应用[J].气象,2007,33(5):20-26.Xiong Q F,Gu Y G,Wang L. Application of SVM method to cloud amount forecast[J]. Meteorological Monthly,2007,33(5):20-26.
[18]吴波,胡邦辉,王学忠,等.基于近似支持向量机的能见度释用预报研究[J].热带气象学报,2017,33(1):104-110.Wu B,Hu B H,Wang X Z,et al. Visibility forecast based on proximal support vector machine[J]. Journal of Tropical Meteorology,2017,33(1):104-110.
[19]郑朝霞,周梅,季致建,等. SVM方法在霾识别和能见度预报中的应用[J].气象科技进展,2016,6(6):30-34.Zheng Z X,Zhou M,Ji Z J,et al. Application of SVM method to identification of haze and prediction of visibility[J]. Advances in Meteorological Science and Technology,2016,6(6):30-34.
[20] Canu S,Rakotomamonjy A. Ozone peak and pollution forecasting using support vectors[R]. Yokohama:IFAC,2001.
[21] García Nieto P J,García-Gonzalo E,Sánchez A B,et al. Air quality modeling using the PSO-SVM-based approach, MLP neural network,and M5 model tree in the metropolitan area of Oviedo(Northern Spain)[J]. Environmental Modeling&Assessment,2018,23(3):229-247.
[22] García Nieto P J,Sánchez Lasheras F,García-Gonzalo E,et al.PM10concentration forecasting in the metropolitan area of Oviedo(Northern Spain)using models based on SVM,MLP,VARMA and ARIMA:a case study[J]. Science of the Total Environment,2018,621:753-761.
[23] Yeganeh B,Motlagh M S P,Rashidi Y,et al. Prediction of CO concentrations based on a hybrid partial least square and support vector machine model[J]. Atmospheric Environment,2012,55:357-365.
[24]朱佳,王振会,金天力,等.基于小波分解和最小二乘支持向量机的大气臭氧含量时间序列预测[J].气候与环境研究,2010,15(3):295-302.Zhu J,Wang Z H,Jin T L,et al. Combination of wavelet decomposition and least square support vector machine to forcast atmospheric ozone content time series[J]. Climatic and Environmental Research,2010,15(3):295-302.
[25] Niu M F,Gan K,Sun S L,et al. Application of decompositionensemble learning paradigm with phase space reconstruction for day-ahead PM2. 5concentration forecasting[J]. Journal of Environmental Management,2017,196:110-118.
[26] Feng Y,Zhang W F,Sun D Z,et al. Ozone concentration forecast method based on genetic algorithm optimized back propagation neural networks and support vector machine data classification[J]. Atmospheric Environment,2011,45(11):1979-1985.
[27] Xu Y Z,Yang W D,Wang J Z. Air quality early-warning system for cities in China[J]. Atmospheric Environment,2017,148:239-257.
[28]王俊秀,安俊琳,邵平,等.南京北郊大气臭氧周末效应特征分析[J].环境科学,2017,38(6):2256-2263.Wang J X,An J L,Shao P,et al. Characteristic study on the“weekend effect” of atmospheric O3in northern Suburb of Nanjing[J]. Environmental Science,2017,38(6):2256-2263.
[29] An J L,Zhu B,Wang H L,et al. Characteristics and source apportionment of VOCs measured in an industrial area of Nanjing, Yangtze River Delta, China[J]. Atmospheric Environment,2014,97:206-214.
[30] Luna A S,Paredes M L L,de Oliveira G C G,et al. Prediction of ozone concentration in tropospheric levels using artificial neural networks and support vector machine at Rio de Janeiro,Brazil[J]. Atmospheric Environment,2014,98:98-104.
[31]安俊琳,朱彬,李用宇.南京北郊大气VOCs体积分数变化特征[J].环境科学,2013,34(12):4504-4512.An J L,Zhu B,Li Y Y. Variation characteristics of ambient volatile organic compounds(VOCs)in Nanjing northern suburb,China[J]. Environmental Science,2013,34(12):4504-4512.
[32] Chelani A B. Prediction of daily maximum ground ozone concentration using support vector machine[J]. Environmental Monitoring and Assessment,2010,162(1-4):169-176.
[33]张玉欣,安俊琳,王俊秀,等.南京工业区挥发性有机物来源解析及其对臭氧贡献评估[J].环境科学,2018,39(2):502-510.Zhang Y X,An J L,Wang J X,et al. Source analysis of volatile organic compounds in the Nanjing industrial area and evaluation of their contribution to ozone[J]. Environmental Science,2018,39(2):502-510.
[34]邵平,安俊琳,杨辉,等.南京北郊夏季近地层臭氧及其前体物体积分数变化特征[J].环境科学,2014,35(11):4031-4043.Shao P,An J L,Yang H,et al. Variation characteristics of surface ozone and its precursors during summertime in Nanjing northern suburb[J]. Environmental Science,2014,35(11):4031-4043.