Ding, Peng (Tjianjin University), Song, Kang (TianJin University), Xie, Hui (Tianjin University), Li, Qian (Tianjin University), Zhou, Shengkai (Guangxi Yuchai Machinery Co., Ltd), Sang, Hailang (Guangxi Yuchai Machinery Co., Ltd)

NOx Emission Prediction Model for Diesel Engines Based on Physical-Informed Data-Driven Ensemble Learning

Scheduled for presentation during the Regular session "Powertrain Control I" (FrA1), Friday, November 1, 2024, 09:30−09:50, Room T1

7th IFAC Conference on Engine and Powertrain Control, Simulation and Modeling, Oct 30 - Nov 1, 2024, Dalian, China

This information is tentative and subject to change. Compiled on January 2, 2025

Abstract

Accurate estimation of nitrogen oxides (NOx) emissions is crucial for post emission treatment of diesel engines, but due to the dynamic operating conditions and complex combustion process, the prediction faces significant challenges. Therefore, this paper proposes a physical-informed data-driven model (PDM) assisted by ensemble learning (EL) to predict NOx emissions. Firstly, a physical laws based prediction model and a feed-forward neural network (FFNN) based prediction model are proposed. Then, physical loss and data loss are combined into the FFNN based model as fusion loss to identify neural network parameters. Next, using EL methods, multiple models are fused to improve their prediction accuracy and generalization ability. 160 data points are collected under steady-state conditions on the diesel engine bench, of which 80% are used for model training and 20% for verification. The results show that the correlation coefficient (R2) of the EL model is improved by 0.72% compared to the PDM, 3.94% compared to the physical model, and 4.00% compared to the FFNN model.