Abstract: Heavy-duty diesel vehicles are important sources of urban nitrogen oxides (NO_x) in actual applications for environmental compliance, emitting more than 80% of NO_x and more than 90% of particulate matter (PM) in total vehicle emissions. The detection and control of heavy-duty diesel emissions are critical for protecting public health. Currently, vehicles on the road must be regularly tested, every six months or once a year, to filter out high-emission mobile sources at vehicle inspection stations. However, it is difficult to effectively screen high-emission vehicles in time with a long interval between annual inspections, and the fixed threshold cannot adapt to the dynamic changes of vehicle driving conditions. An on-board diagnostic device (OBD) is installed inside the vehicle and can record the vehicle’s emission data in real time. In this paper, we propose a temporal optimization long short-term memory (LSTM) and adaptive dynamic threshold approach to identify heavy-duty high-emitters by using OBD data, which can continuously track and record the emission status in real time. First, a temporal optimization LSTM emission prediction model is established to solve the attention bias discrepancy problem on time steps that is caused by the large number of OBD data streams in practice. Then, the concentration prediction error sequence is detected and distinguished from the anomalous emission contexts using flexible criteria, calculated by an adaptive dynamic threshold with changing driving conditions. Finally, a similarity metric strategy for the time series is introduced to correct some pseudo anomalous results. Experiments on three real OBD time-series emission datasets demonstrate that our method can achieve high accuracy anomalous emission identification.

基于时序优化长短期记忆和自适应阈值的高排放重型柴油车识别

[1]Chandola V, Banerjee A, Kumar V, 2009. Anomaly detection: a survey. ACM Comput Surv, 41(3):15.

[2]Cho K, van Merriënboer B, Gulcehre C, et al., 2014. Learning phrase representations using RNN encoder-decoder for statistical machine translation. Proc Conf on Empirical Methods in Natural Language Processing, p.1724-1734.

[3]Fang SW, Li Q, Karimian H, et al., 2022. DESA: a novel hybrid decomposing-ensemble and spatiotemporal attention model for PM_2.5 forecasting. Environ Sci Poll Res, 29(36):54150-54166.

[4]Franco V, Kousoulidou M, Muntean M, et al., 2013. Road vehicle emission factors development: a review. Atmos Environ, 70:84-97.

[5]Guo HF, Zeng J, Hu YM, 2006. Neural network modeling of vehicle gross emitter prediction based on remote sensing data. Proc IEEE Int Conf on Networking, Sensing and Control, p.943-946.

[6]He ZY, Xu XF, Deng SC, 2003. Discovering cluster-based local outliers. Patt Recogn Lett, 24(9-10):1641-1650.

[7]Jiang MF, Tseng SS, Su CM, 2001. Two-phase clustering process for outliers detection. Patt Recogn Lett, 22(6-7):691-700.

[8]Karimian H, Li Q, Li CC, et al., 2019. Spatio-temporal variation of wind influence on distribution of fine particulate matter and its precursor gases. Atmos Poll Res, 10(1):53-64.

[9]Li YR, Zhu ZF, Kong DQ, et al., 2019. EA-LSTM: evolutionary attention-based LSTM for time series prediction. Knowl-Based Syst, 181:104785.

[10]Li ZR, Kang Y, Lv WJ, et al., 2021. High-emitter identification model establishment using weighted extreme learning machine and active sampling. Neurocomputing, 441:79-91.

[11]Liu FT, Ting KM, Zhou ZH, 2008. Isolation forest. Proc 8^th IEEE Int Conf on Data Mining, p.413-422.

[12]Liu YQ, Gong CY, Yang L, et al., 2020. DSTP-RNN: a dual-stage two-phase attention-based recurrent neural network for long-term and multivariate time series prediction. Expert Syst Appl, 143:113082.

[13]Lucas JM, Saccucci MS, 1990. Exponentially weighted moving average control schemes: properties and enhancements. Technometrics, 32(1):1-12.

[14]Lukashevich H, Nowak S, Dunker P, 2009. Using one-class SVM outliers detection for verification of collaboratively tagged image training sets. Proc IEEE Int Conf on Multimedia and Expo, p.682-685.

[15]Malhotra P, Vig L, Shroff GM, et al., 2015. Long short term memory networks for anomaly detection in time series. Proc 23^rd European Symp on Artificial Neural Networks, Computational Intelligence and Machine Learning.

[16]McClintock PM, 2007. High Emitter Remote Sensing Project. Prepared for Southeast Michigan Council of Governments. http://refhub.elsevier.com/S1352-2310(18)30187-0/sref52 [Accessed on Mar. 29, 2022].

[17]McClintock PM, 2011. The Colorado Remote Sensing Program January–December 2010. The Colorado Department of Public Health and Environment. http://refhub.elsevier.com/S1352-2310(18)30187-0/sref80 [Accessed on Mar. 29, 2022].

[18]Ministry of Ecology and Environment of the People’s Republic of China, 2022. China Mobile Source Environmental Management Annual Report (in Chinese). https://www.mee.gov.cn/hjzl/sthjzk/ydyhjgl/202212/W020221207387013521948.pdf [Accessed on Mar. 29, 2022].

[19]Pujadas M, Domínguez-Sáez A, de la Fuente J, 2017. Real-driving emissions of circulating Spanish car fleet in 2015 using RSD technology. Sci Total Environ, 576:193-209.

[20]Senin P, 2008. Dynamic Time Warping Algorithm Review. University of Hawaii, Honolulu, USA.

[21]Shipmon DT, Gurevitch JM, Piselli PM, et al., 2017. Time series anomaly detection; detection of anomalous drops with limited features and sparse examples in noisy highly periodic data. https://arxiv.org/abs/1708.03665

[22]Smit R, Bluett J, 2011. A new method to compare vehicle emissions measured by remote sensing and laboratory testing: high-emitters and potential implications for emission inventories. Sci Total Environ, 409(13):2626-2634.

[23]Stephens RD, Cadle SH, Qian TZ, 1996. Analysis of remote sensing errors of omission and commission under FTP conditions. J Air Waste Manag Assoc, 46(6):510-516.

[24]Williamson DF, Parker RA, Kendrick JS, 1989. The box plot: a simple visual method to interpret data. Ann Int Med, 110(11):916-921.

[25]Wu CL, Li Q, Hou JX, et al., 2018. PM_2.5 concentration prediction using convolutional neural networks. Sci Surv Map, 43(8):68-75 (in Chinese).

[26]Xie H, Zhang YJ, He Y, et al., 2019. Automatic and fast recognition of on-road high-emitting vehicles using an optical remote sensing system. Sensors, 19(16):3540.

[27]Xie H, Zhang YJ, He Y, et al., 2021. Parallel attention-based LSTM for building a prediction model of vehicle emissions using PEMS and OBD. Measurement, 185:110074.

[28]Xu XW, Yuruk N, Feng ZD, et al., 2007. SCAN: a structural clustering algorithm for networks. Proc 13^th ACM SIGKDD Int Conf on Knowledge Discovery and Data Mining, p.824-833.

[29]Xu ZY, Kang Y, Cao Y, et al., 2021. Spatiotemporal graph convolution multifusion network for urban vehicle emission prediction. IEEE Trans Neur Netw Learn Syst, 32(8):3342-3354.

[30]Yu Y, Si XS, Hu CH, et al., 2019. A review of recurrent neural networks: LSTM cells and network architectures. Neur Comput, 31(7):1235-1270.

[31]Zeng J, Guo HF, Hu YM, 2008. A PKGV-ANN model for vehicle high emitters identification based on remote sensing data. Proc 27^th Chinese Control Conf, p.171-175.

[32]Zhang GP, 2003. Time series forecasting using a hybrid ARIMA and neural network model. Neurocomputing, 50:159-175.

[33]Zhou HY, Zhang SH, Peng JQ, et al., 2021. Informer: beyond efficient transformer for long sequence time-series forecasting. Proc 35^th AAAI Conf on Artificial Intelligence, p.11106-11115.