﻿ 融合GNSS气象参数的BP神经网络雾霾预测研究
 文章快速检索 高级检索
 大地测量与地球动力学  2019, Vol. 39 Issue (11): 1148-1152  DOI: 10.14075/j.jgg.2019.11.010

### 引用本文

ZHOU Yongjiang, YAO Yibin, YAN Xiao, et al. Study on Haze Prediction of BP Neural Network Incorporating GNSS Meteorological Parameters[J]. Journal of Geodesy and Geodynamics, 2019, 39(11): 1148-1152.

### Foundation support

National Natural Science Foundation of China, No. 41574028.

### Corresponding author

YAO Yibin, PhD, professor, PhD supervisor, majors in GNSS meteorology, E-mail:ybyao@whu.edu.cn.

### About the first author

ZHOU Yongjiang, postgraduate, majors in GNSS tropospheric application, E-mail:1901985897@qq.com.

### 文章历史

1. 武汉大学测绘学院，武汉市珞喻路129号，430079

1 雾霾预测模型 1.1 BP神经网络

BP神经网络(BPNN)是一种按误差反向传播算法训练的多层前馈网络模型，多用于函数逼近、预测分析等[12]，其网络结构通常包含3个层次：输入层、隐含层及输出层。图 1为3层结构的BP神经网络模型结构。

 图 1 3层BP神经网络结构 Fig. 1 The structure of three-layer BPNN

BP神经网络每层由一个或多个神经元(网络节点)构成，神经元输入与输出之间的关系为：

 ${I_j} = \sum\limits_{j = 1}^n {{w_{kj}}} {x_j} - {\theta _k}, {y_k} = f\left( {{I_j}} \right)$ (1)

1.2 雾霾预测模型

 图 2 融合时序网络和回归网络的雾霾预测模型 Fig. 2 The haze prediction model combines time series network and regression network
2 数据处理与分析 2.1 数据采样及模型训练

2.2 [TSA+RA]-ANN、TSA-ANN模型预测性能比较

 图 3 3种分类情况下预测值与实测值的变化对比结果及对应的误差绝对值 Fig. 3 The comparison between predicted value and measured value in three classification cases and the corresponding absolute value of the error

2.3 [TSA+RA]-ANN模型精度评价

 ${\rm{RMSE}} = \sqrt {\frac{1}{N}\sum\limits_{t = 1}^N {{{\left( {{O_t} - {P_t}} \right)}^2}} }$ (2)
 ${\rm{MAD}} = \frac{1}{N}\sum\limits_{t = 1}^N {\left| {{O_t} - {P_t}} \right|}$ (3)
 ${R^2} = {\left( {\frac{1}{{N - 1}}\sum\limits_{t = 1}^N {\left( {\frac{{{O_t} - \bar O}}{{{S_O}}}} \right)} \left( {\frac{{{P_t} - \bar P}}{{{S_P}}}} \right)} \right)^2}$ (4)

3 结语

 [1] Lei Y H. Hazards of Haze and Countermeasures[J]. Applied Mechanics and Materials, 2014, 507: 817-820 DOI:10.4028/www.scientific.net/AMM.507.817 (0) [2] Zhang Q, Quan J N, Tie X X, et al. Effects of Meteorology and Secondary Particle Formation on Visibility during Heavy Haze Events in Beijing, China[J]. Science of the Total Environment, 2015, 502: 578-584 DOI:10.1016/j.scitotenv.2014.09.079 (0) [3] Aldape F, Flores M J, Flores A J, et al. Elemental Composition and Source Identification of PM2.5 Particles Collected in Downtown Mexico City[J]. International Journal of PIXE, 2005, 15(3-4): 263-270 (0) [4] Yang X P, Zhang Z X, Zhang Z Q, et al. A Long-Term Prediction Model of Beijing Haze Episodes Using Time Series Analysis[J]. Computational Intelligence and Neuroscience, 2016 (0) [5] Dong M, Yang D, Kuang Y, et al. PM2.5 Concentration Prediction Using Hidden Semi-Markov Model-Based Times Series Data Mining[J]. Expert Systems with Applications, 2009, 36(5): 9 046-9 055 DOI:10.1016/j.eswa.2008.12.017 (0) [6] Pérez P, Trier A, Reyes J. Prediction of PM2.5, Concentrations Several Hours in Advance Using Neural Networks in Santiago, Chile[J]. Atmospheric Environment, 2000, 34(8): 1 189-1 196 DOI:10.1016/S1352-2310(99)00316-7 (0) [7] Wu Y C, Feng J W. Development and Application of Artificial Neural Network[J]. Wireless Personal Communications, 2018, 102(2): 1 645-1 656 DOI:10.1007/s11277-017-5224-x (0) [8] Specht D F. A General Regression Neural Network[J]. IEEE Transactions on Neural Networks, 1991, 2(6): 568-576 DOI:10.1109/72.97934 (0) [9] Wen X H, Chen K Z. Time Series Neural Network Forecasting Methods[J]. Journal of Electronics, 1996, 12(1): 1-8 (0) [10] Gemello R, Albesano D, Mana F. Multi-Source Neural Networks for Speech Recognition[C]. International Joint Conference on Neural Networks, Washington DC, 1999 (0) [11] Rumerlhar D E, Hinton G E, Williams R J. Learning Representation by Back-Propagating Errors[J]. Cognitive Modeling, 1986, 5(3): 533-536 (0) [12] Goh A T C. Back-Propagation Neural Networks for Modeling Complex Systems[J]. Artificial Intelligence in Engineering, 1995, 9(3): 143-151 (0) [13] Jin W, Li Z J, Wei L S, et al. The Improvements of BP Neural Network Learning Algorithm[C]. 5th International Conference on Signal Processing Proceedings, 16th World Computer Congress, Beijing, 2000 (0) [14] Xu J J, Liu B, Yuan J G, et al. Inversion of Precipitable Water Vapor in Hongkong[C]. International Workshop on Earth Observation and Remote Sensing Applications, Beijing, 2008 (0) [15] HJ 633-2012, 环境空气质量指数(AQI)技术规定(试行)[S].北京: 中国环境科学出版社, 2016 (HJ 633-2012, Environmental Air Quality Index (AQI) Technical Regulations (Trial)[S]. Beijing: China Environmental Science Press, 2016) (0) [16] Willmott C J, Ackleson S G, Davis R E, et al. Statistics for the Evaluation and Comparison of Models[J]. Journal of Geophysical Research: Oceans, 1985, 90(C5): 8 995-9 005 DOI:10.1029/JC090iC05p08995 (0) [17] Gunhan T, Demir V, Hancioglu E, et al. Mathematical Modeling of Drying of Bay Leaves[J]. Energy Conversion and Management, 2005, 46(11-12): 1 667-1 679 DOI:10.1016/j.enconman.2004.10.001 (0) [18] Nastos P T, Paliatsos A G, Koukouletsos K V, et al. Artificial Neural Networks Modeling for Forecasting the Maximum Daily Total Precipitation at Athens, Greece[J]. Atmospheric Research, 2014, 144: 141-150 DOI:10.1016/j.atmosres.2013.11.013 (0) [19] Ni X L, Cao C X, Zhou Y K, et al. Spatio-Temporal Pattern Estimation of PM2.5 in Beijing-Tianjin-Hebei Region Based on MODIS AOD and Meteorological Data Using the Back Propagation Neural Network[J]. Atmosphere, 2018, 9(3): 105 DOI:10.3390/atmos9030105 (0) [20] Zhao X R, Shi H Q, Yu H, et al. Inversion of Nighttime PM2.5 Mass Concentration in Beijing Based on the VⅡRS Day-Night Band[J]. Atmosphere, 2016, 7(10): 136 DOI:10.3390/atmos7100136 (0) [21] Feng X, Li Q, Zhu Y J, et al. Artificial Neural Networks Forecasting of PM2.5Pollution Using Air Mass Trajectory Based Geographic Model and Wavelet Transformation[J]. Atmospheric Environment, 2015, 107: 118-128 DOI:10.1016/j.atmosenv.2015.02.030 (0)
Study on Haze Prediction of BP Neural Network Incorporating GNSS Meteorological Parameters
ZHOU Yongjiang1     YAO Yibin1     YAN Xiao1     ZHAO Cunjie1
1. School of Geodesy and Geomatics, Wuhan University, 129 Luoyu Road, Wuhan, 430079, China
Abstract: Base on the meteorological parameters (temperature (T), air pressure (P), and precipitable water vapor (PWV)), of Beijing Fangshan Station released by the IGS Center and PM2.5 data for the same period, this paper establishes a haze prediction model combining time series network and regression network to predict PM2.5 concentration. The research shows that the fusion network model introducing GNSS meteorological parameters is more adaptable and accurate than the single network model, that it can accurately predict the change of PM2.5 within a certain accuracy range, and that timeliness can reach 3h. Related studies have verified the feasibility of satellite navigation technology for monitoring and forecasting of haze weather.
Key words: haze; GNSS tropospheric parameters; BP neural network; fusion network model