基于多模态支持向量回归的PM_(2.5)浓度预测
|
摘要
为更好地掌握日均PM_(2.5)浓度的变化规律,提出了一种基于多模态支持向量回归(MSVR)的混合预测模型。利用集成经验模态分解将日均PM_(2.5)数据分解成不同频段的分量序列,以降低数据的非平稳性。然后根据每组分量自身特点构建不同的支持向量回归(SVR)模型,并通过相关分析确定各分量输入变量。最后,将各分量预测值进行叠加得到最终预测结果。以浙江省玉环市的PM_(2.5)浓度进行验证。结果表明:与单一SVR模型相比,MSVR模型具有更好的预测效果,精度评价指标MAE、MAPE和RMSE分别下降了26.98%、23.04%、34.08%,这为大气污染预控提供了有效的技术支持。
In order to better grasp the change rule of average daily PM_(2.5) concentration, a hybrid forecasting model based on multimodal support vector regression(MSVR) was proposed. Firstly, average daily data of PM_(2.5) concentration was decomposed into several scales with different frequency bands by ensemble empirical mode decomposition(EEMD) to decrease the data non-stationary. Secondly, at each scale, appropriate forecasting model was built using support vector regression(SVR) in the light of its own characteristics, and the input variables of each scale were also determined by correlation analysis. Finally, the forecasting value of each scale was superimposed to gain the final result. The experimental confirmation of the proposed model was tested by applying PM_(2.5) concentration data in Yuhuan City, Zhejiang Province. The results indicated that the MSVR model outperformed the single SVM model, with the accuracy evaluation indexes MAE, MAPE and RMSE decreased by 26.98%, 23.04% and 34.08% respectively, which provided effective technical support for pre-control of air pollution.
In order to better grasp the change rule of average daily PM_(2.5) concentration, a hybrid forecasting model based on multimodal support vector regression(MSVR) was proposed. Firstly, average daily data of PM_(2.5) concentration was decomposed into several scales with different frequency bands by ensemble empirical mode decomposition(EEMD) to decrease the data non-stationary. Secondly, at each scale, appropriate forecasting model was built using support vector regression(SVR) in the light of its own characteristics, and the input variables of each scale were also determined by correlation analysis. Finally, the forecasting value of each scale was superimposed to gain the final result. The experimental confirmation of the proposed model was tested by applying PM_(2.5) concentration data in Yuhuan City, Zhejiang Province. The results indicated that the MSVR model outperformed the single SVM model, with the accuracy evaluation indexes MAE, MAPE and RMSE decreased by 26.98%, 23.04% and 34.08% respectively, which provided effective technical support for pre-control of air pollution.
引文
[1] 吴剑斌, 王茜, 伏晴艳. 2013年12月上海市PM2.5重污染过程数值模拟研究[J]. 环境科学学报, 2016, 36(6):2152-2159.
[2] Zhang H, Denero S P, Joe D K, et al. Development of a source oriented version of the WRF/Chem model and its application to the California regional PM10/PM2.5 air quality study[J]. Atmospheric Chemistry and Physics, 2014, 14(1):485-503.
[3] Vlachogianni A, Kassomenos P, Karppinen A, et al. Evaluation of a multiple regression model for the forecasting of the concentrations of NOx, and PM10, in Athens and Helsinki[J]. Science of the Total Environment, 2011, 409(8):1559-1571.
[4] Singh K P, Gupta S, Kumar A, et al. Linear and nonlinear modeling approaches for urban air quality prediction[J]. Science of the Total Environment, 2012, 426(2):244-255.
[5] 郭飞, 谢立勇. 基于气象因素和改进支持向量机的空气质量指数预测[J]. 环境工程, 2017, 35(10):151-155.
[6] 荆涛, 李霖, 于文柱, 等. t分布受控遗传算法优化BP神经网络的PM2.5质量浓度预测[J]. 中国环境监测, 2015, 31(4):100-105.
[7] Feng X, Li Q, Zhu Y, et al. Artificial neural networks forecasting of PM2.5, pollution using air mass trajectory based geographic model and wavelet transformation[J]. Atmospheric Environment, 2015, 107:118-128.
[8] Li W, Kong D, Wu J. A new hybrid model FPA-SVM considering cointegration for particular matter concentration forecasting: a case study of Kunming and Yuxi, China[J]. Computational Intelligence and Neuroscience, 2017, 2017: 2843651.
[9] 刘杰, 杨鹏, 吕文生, 等. 模糊时序与支持向量机建模相结合的PM2.5质量浓度预测[J]. 北京科技大学学报, 2014, 36(12):1694-1702.
[10] 秦喜文, 刘媛媛, 王新民, 等. 基于整体经验模态分解和支持向量回归的北京市PM2.5预测[J]. 吉林大学学报(地球科学版), 2016, 46(2):563-568.
[11] Wang W C, Chau K W, Xu D M, et al. Improving forecasting accuracy of annual runoff time series using ARIMA based on EEMD decomposition[J]. Water Resources Management, 2015, 29(8):2655-2675.
[12] Suykens J A K, Gestel T V, Brabanter J D, et al. Least squares support vector machines[M]. Singapore: World Scientific, 2002.
[13] 喻其炳, 李勇, 白云, 等. 基于聚类分析与偏最小二乘法的支持向量机PM2.5预测[J]. 环境科学与技术, 2017, 40(6):157-164.
[14] Wang S, Yu L, Tang L, et al. A novel seasonal decomposition based least squares support vector regression ensemble learning approach for hydropower consumption forecasting in China[J]. Energy, 2011, 36(11):6542-6554.
[15] 李勇, 白云, 李川. 基于小波分析与BP神经网络的PM10浓度预测模型[J]. 环境监测管理与技术, 2016, 28(5):24-28.
[16] Bai Y, Li Y, Wang X, et al. Air pollutants concentrations forecasting using back propagation neural network based on wavelet decomposition with meteorological conditions[J]. Atmospheric Pollution Research, 2016, 7(3):557-566.
[2] Zhang H, Denero S P, Joe D K, et al. Development of a source oriented version of the WRF/Chem model and its application to the California regional PM10/PM2.5 air quality study[J]. Atmospheric Chemistry and Physics, 2014, 14(1):485-503.
[3] Vlachogianni A, Kassomenos P, Karppinen A, et al. Evaluation of a multiple regression model for the forecasting of the concentrations of NOx, and PM10, in Athens and Helsinki[J]. Science of the Total Environment, 2011, 409(8):1559-1571.
[4] Singh K P, Gupta S, Kumar A, et al. Linear and nonlinear modeling approaches for urban air quality prediction[J]. Science of the Total Environment, 2012, 426(2):244-255.
[5] 郭飞, 谢立勇. 基于气象因素和改进支持向量机的空气质量指数预测[J]. 环境工程, 2017, 35(10):151-155.
[6] 荆涛, 李霖, 于文柱, 等. t分布受控遗传算法优化BP神经网络的PM2.5质量浓度预测[J]. 中国环境监测, 2015, 31(4):100-105.
[7] Feng X, Li Q, Zhu Y, et al. Artificial neural networks forecasting of PM2.5, pollution using air mass trajectory based geographic model and wavelet transformation[J]. Atmospheric Environment, 2015, 107:118-128.
[8] Li W, Kong D, Wu J. A new hybrid model FPA-SVM considering cointegration for particular matter concentration forecasting: a case study of Kunming and Yuxi, China[J]. Computational Intelligence and Neuroscience, 2017, 2017: 2843651.
[9] 刘杰, 杨鹏, 吕文生, 等. 模糊时序与支持向量机建模相结合的PM2.5质量浓度预测[J]. 北京科技大学学报, 2014, 36(12):1694-1702.
[10] 秦喜文, 刘媛媛, 王新民, 等. 基于整体经验模态分解和支持向量回归的北京市PM2.5预测[J]. 吉林大学学报(地球科学版), 2016, 46(2):563-568.
[11] Wang W C, Chau K W, Xu D M, et al. Improving forecasting accuracy of annual runoff time series using ARIMA based on EEMD decomposition[J]. Water Resources Management, 2015, 29(8):2655-2675.
[12] Suykens J A K, Gestel T V, Brabanter J D, et al. Least squares support vector machines[M]. Singapore: World Scientific, 2002.
[13] 喻其炳, 李勇, 白云, 等. 基于聚类分析与偏最小二乘法的支持向量机PM2.5预测[J]. 环境科学与技术, 2017, 40(6):157-164.
[14] Wang S, Yu L, Tang L, et al. A novel seasonal decomposition based least squares support vector regression ensemble learning approach for hydropower consumption forecasting in China[J]. Energy, 2011, 36(11):6542-6554.
[15] 李勇, 白云, 李川. 基于小波分析与BP神经网络的PM10浓度预测模型[J]. 环境监测管理与技术, 2016, 28(5):24-28.
[16] Bai Y, Li Y, Wang X, et al. Air pollutants concentrations forecasting using back propagation neural network based on wavelet decomposition with meteorological conditions[J]. Atmospheric Pollution Research, 2016, 7(3):557-566.