Parallel attention-based LSTM for building a prediction model of vehicle emissions using PEMS and OBD

• Published in: Measurement : journal of the International Measurement Confederation Vol. 185; p. 110074
• Main Authors: Xie, Hao, Zhang, Yujun, He, Ying, You, Kun, Fan, Boqiang, Yu, Dongqi, Lei, Boen, Zhang, Wangchun
• Format: Journal Article
• Language: English
• Published: London Elsevier Ltd 01-11-2021; Elsevier Science Ltd
• Subjects: Decoding; Deep learning; Early warning of high emissions; Electric vehicles; Emission analysis; Emission standards; Emissions; Emissions control; Experiments; Measurement methods; Outliers (statistics); Outliers elimination; PEMS; Prediction models; Predictions; Road tests; Time series prediction; Vehicle emissions; Deep learning; Time series prediction; Early warning of high emissions; Outliers elimination; PEMS
• ISSN: 0263-2241; 1873-412X
• DOI: 10.1016/j.measurement.2021.110074

•DL combined with PEMS and OBD for building prediction model of vehicle emissions.•An innovative attention encoding structure was proposed to accelerate convergence.•An automatic elimination method for the outliers in the PEMS data was proposed.•An early warning system of high vehicle emissions was proposed.•A model that can be trained with lab tests to predict on-road results was proposed.
Portable emission measurement system (PEMS) testing, which is the most accurate measurement method for vehicle emissions, has been included into the regulations of vehicle emission standards in various countries. However, PEMS is expensive, and in the long-term measurement process, the monitoring data will exhibit outliers, which is a drift phenomenon. In addition, as a measurement method, it cannot prevent the occurrence of high vehicle emissions. To solve the above problem, this study proposes a parallel attention-based long short-term memory (PA-LSTM) for building an emission prediction model using PEMS and on board diagnostics (OBD). According to the characteristics of the real vehicle road test and bench test data, the PA-LSTM model adopts a parallel spatial attention coding mechanism, combined with a temporal attention decoding mechanism. Qualitative and quantitative experimental results show that the PA-LSTM model can achieve a more accurate prediction of vehicle emissions compared with other popular models, and the proposed model can eliminate outliers and restrain the offset of the zero levels in the PEMS data. The most significant thing is that the PA-LSTM model can foresee about the possible high vehicle emissions in the future and provide timely feed-back to the emission control system of the vehicle engine, so as to make corresponding control measurements in time and avoid the occurrence of high emissions.