PREDICTION OF AIR POLLUTANT CONCENTRATIONS IN XUZHOU BASED ON CEEMD-BiGRU MODEL
-
摘要: 目前大气污染物对于地区经济以及人体健康的影响不容忽视。选取徐州市2016-01-01—2021-01-24大气污染物和气象要素数据,针对大气污染物浓度波动性强等特点,运用互补集成经验模态分解(CEEMD)将污染物数据分解为本征模态分量,提取出原始数据的各项特征,再对分解出的各本征模态分量构建双向门控循环单元模型(BiGRU),通过双向循环训练,学习各分量的特征趋势并获得最优训练参数,将输出结果重构,得到最终的预测值。结果表明:与BiGRU、BP模型相比,CEEMD-BiGRU模型预测各项大气污染物的平均绝对误差(MAE)、均方根误差(RMSE)和平均绝对百分比误差(MAPE)分别下降15%、20%和2百分点以上,预测精度有较大提升。在此基础上,利用CEEMD-BiGRU模型预测后一时间段残差,以修正原预测值,得到大气污染物预测区间上界,进一步扩展模型的适用性。
-
关键词:
- 大气污染物预测 /
- 互补集成经验模态分解 /
- 双向门控循环单元 /
- 区间预测
Abstract: The current impact of air pollutants on the regional economy and human health cannot be ignored. This paper selected data on atmospheric pollutants and meteorological elements in Xuzhou from January 1, 2016 to January 24, 2021. Given the characteristics of strong fluctuation of atmospheric pollutant concentration, the pollutant data were decomposed into intrinsic mode fuction by using complementary ensemble empirical mode decomposition(CEEMD), and various features of the original data were extracted. Each decomposed intrinsic mode fuction was used as the input layer of the bidirectional gated recurrent model(BiGRU). Through bidirectional cyclic training, the characteristic trend of each component was learned and the optimal training parameters were obtained. The output results of each intrinsic mode fuction were reconstructed to obtain the final predicted value. The results showed that compared with the BiGRU and BP models, MAE, RMSE and MAPE of each air pollutant predicted by CEEMD-BiGRU model were reduced by more than 15%, 20% and 2 percentage points, and the prediction accuracy was greatly improved. On this basis, CEEMD-BiGRU model was used to predict the residuals of the latter period to correct the original prediction values and obtain the upper bound of the prediction interval for air pollutants, further to extend the applicability of the model. -
[1] 毕谆,周红刚.PM2.5引起的肺部疾病及其作用机制的研究进展[J].环境工程,2016,34(增刊1):496-499. [2] 谢元博,陈娟,李巍.雾霾重污染期间北京居民对高浓度PM2.5持续暴露健康风险及其损害价值评估[J].环境科学,2014,35(1):1-8. [3] LI X L,WANG Y Q,ZHANG Y,et al.Response of soil chemical properties and enzyme activity of four species i-n the Three Gorges Reservoir area to simulated acid rain[J].Ecotoxicology and Environmental Safety,2021,208:111457. [4] BARRECA A I,NEIDELL M,SANDERS N J.Long-run pollution exposure and mortality:evidence from the acid rain program[J].Journal of Public Economics,2021,200:104440. [5] LI M M,DONG H,WANG B G,et al.Association between ambient ozone pollution and mortality from a s-pectrum of causes in Guangzhou,China[J].Science of the Total Environment,2021,754:142110. [6] THONGTHAMMACHART T,ARAKI S,SHIMADERA H,et al.An integrated model combining random forests and WRF/C-MAQ model for high accuracy spatiotemporal PM2.5 predictions in the Kansai region of Japan[J].Atmospheric Environment,2021,262:118620. [7] CHARU S,SANJEEV K S,PRAKASH C,et al.Simulation of an extreme dust episode using WRF-CHEM based on optimal ensemble approach[J].Atmospheric Research,2021,249:105296. [8] 周广强,高伟,谷怡萱,等.WRF-Chem模式降水对上海PM2.5预报的影响[J].环境科学学报,2017,37(12):4476-4482. [9] WANG X H,WANG B Z.Research on prediction of environmental aerosol and PM2.5 based on artificial neural network[J].Neural Computing and Applications,2019,31(12):8217-8227. [10] CHAUDHURI S,DUTTA D.Mann-Kendall trend of pollutants,temperature and humidity over an urban stat-ion of India with forecast verification using different ARIMA models[J].Environmental Monitoring and Assessment,2014,186(8):4719-4742. [11] TAO Q,LIU F,LI Y,et al.Air pollution forecasting using a deep learning model based on 1D convnetsand bidirectional GRU[J].IEEE Access,2019,7:76690-76698. [12] FENG R,ZHENG H J,GAO H,et al.Recurrent neural network and random forest for analysis and accurate forecast of atmospheric pollutants:a case study in Hangzhou,China[J].Journal of Cleaner Production,2019,231:1005-1015. [13] 成华义.GA-PSO-BP神经网络在大气污染物浓度预测中的应用研究[D].武汉:华中科技大学,2014. [14] 郑霞,胡东滨,李权.基于小波分解和SVM的大气污染物浓度预测模型研究[J].环境科学学报,2020,40(8):2962-2969. [15] 秦喜文,王强进,王新民,等.基于VMD和LSTM方法的北京市PM2.5短期预测[J].吉林大学学报(地球科学版),2022,52(1):214-221. [16] 赵彦明.基于时空相关性的LSTM算法及PM2.5浓度预测应用[J].计算机应用与软件,2021,38(6):249-255,323. [17] SCHUSTER M,PALIWAL K K.Bidirectional recurrent neural networks[J].IEEE transactions on Signal Processing,1997,45(11):2673-2681. [18] CAO X K,REN N,TIAN G L,et al.A three-dimensional prediction method of dissolved oxygen in pond culture based on Attention-GRU-GBRT[J].Computers and Electronics in Agriculture,2021,181:105955. [19] LIANG R,CHANG X T,JIA P T,et al.Mine gas concentration forecasting model based on an optimized BiGRU network[J].ACS Omega,2020,5(44):28579-28586. [20] KONSTANTIN Y V,RUSSELL R D,NICKOLAY A K,et al.The net decay time of anomalies in concentrations of atmospheric pollutants[J].Atmospheric Environment,2017,160:19-26. [21] GU J,PENG Y X.An improved complementary ensemble empirical mode decomposition method and its application in rolling bearing fault diagnosis[J].Digital Signal Processing,2021,113:103050. [22] 陈煜升,张宇静,赵天良,等.近5年徐州市大气污染变化及相关气象作用[J].环境科学与技术,2019,42(增刊1):152-158. [23] ABDOLLAH B,OSMAN E O,DEVIN K H.Structural system identification based on variational mode decomposition[J].Journal of Sound and Vibration,2018,417:182-197. [24] THIRUMALAISAMY M R,ANSELL P J.Fast and adaptive empirical mode decomposition for multidimensional,multivariate signals[J].IEEE Signal Processing Letters,2018,25(10):1550-1554. [25] REZAEI H,FAALJOU H,MANSOURFAR G.Stock price prediction using deep learning and frequency decomposition[J].Expert Systems with Applications,2021,169:114332. [26] CHEN J X,FENG X,JIANG L,et al.State of charge estimation of lithium-ion battery using denoising autoe-noder and gated recurrent unit recurrent neural network[J].Energy,2021,227:120451. [27] LONG D K,ZHANG R C,MAO Y Y.Prototypical recurrent unit[J].Neurocomputing,2018,311:146-154. [28] LI P,WANG S W,JI H,et al.Air Quality index prediction based on an adaptive dynamic particle swarm optimized bidirectional gated recurrent neural network-China region[J].Advanced Theory and Simulations,2021,4(12):2100220-1-2100220-14. [29] NITISH S,GEOFFREY E H,ALEX K,et al.Dropout:a simple way to prevent neural networks from overfitting[J].Journal of Machine Learning Research,2014,15(1):1929-1958. [30] MAHMOOD D,MOHAMMAD H S.EclatDS:an efficient sliding window based frequent pattern miningmethod for data streams[J].Intelligent Data Analysis,2011,15(4):571-587. [31] SHRINIVAS C,PETER J,SANDER W.Fast nonlinear fourier transform algorithms using higher order exponential integrators[J].IEEE Access,2019,7:145161-145176. [32] DIEDERIK P K,JIMMY B.Adam:a method for stochastic optimization[J].Computer Science,2014.
点击查看大图
计量
- 文章访问数: 346
- HTML全文浏览量: 53
- PDF下载量: 21
- 被引次数: 0