Air.Car

Air.Car - Using artificial neural network to predict NOx emissions in real-time

Despite the introduction of stringent emission and air quality regulations, NO₂ concentrations are still exceeding harmful levels in a large number of cities across Europe. Discrepancies between standardized type-approval and real-world driving emissions [1], which partially explain the failure to fulfill air quality European regulations, led the introduction of the Real Driving Emission (RDE) regulation in September 2017. In parallel, there is an increasing wish from authorities to tax fairly drivers entering polluted areas based on their actual emissions rather than on the Euro standard or other type of metric that does not account for the influence of individual driving behaviour or the variability across the manufacturers.

In the previously described context, existing emission models are not able to tackle the problem: physic-based models are computationally expensive and require very often to be calibrated, while lookup table or average-speed (such as COPERT) models are not able to capture accurately highly non-linear behaviours associated with NO_x emissions [2]. There is consequently a need to develop new emission models, able at running in real-time at a low computational cost while still providing the level of accuracy required to implement a polluter-tax system based on their output. With the increase in computational power and availability of real-world driving data, the potential of Artificial Neural Networks (ANNs) to address this issue cannot be denied.

While simple ANNs architectures have been used to predict NO_x emissions for selected vehicles and specific conditions [3], this study uses a large dataset of diesel vehicles (passenger cars, buses) tested in real-world conditions to develop a real-time ANN NOx emission model. Vehicle driving parameters (engine speed, engine load, mass air flow, etc.) used as inputs to the model were obtained from the On-Board Diagnostic device (OBD-II) while emissions were measured with a Portable Emission Measurement System (PEMS).

Research team:

• Dr. Marc Stettler
• Clémence Le Cornec
• Dr. Mino Wop

References:

[1] Rosalind O'Driscoll, Marc E.J. Stettler, Nick Molden, Tim Oxley, Helen M. ApSimon, “Real world CO2 and NOx emissions from 149 Euro 5 and 6 diesel, gasoline and hybrid passenger cars”, Science of The Total Environment, Volume 621, 2018, Pages 282-290.

[2] J. M. Alonso, F. Alvarruiz, J. M. Desantes, L. Hernandez, V. Hernandez and G. Molto, "Combining Neural Networks and Genetic Algorithms to Predict and Reduce Diesel Engine Emissions," in IEEE Transactions on Evolutionary Computation, vol. 11, no. 1, pp. 46-55, Feb. 2007.

[3] Obodeh, O., and C. I. Ajuwa. "Evaluation of artificial neural network performance in predicting diesel engine NOx emissions." European Journal of Scientific Research 33.4 (2009): 642-653.