NOX Emission Prediction of Diesel Engine Based on GWO-LSTM

Abstract

Diesel engine NOx is the main harmful emission of motor vehicles. Accurate measurement of NOx emission is beneficial to the control of SCR (selective catalytic reduction) urea injection so as to reduce emissions. At present, NOx emission value is mainly obtained by NOx sensor or MAP calibration and these two methods have limitations in practical applications. In this study, PCA (principal component analysis) is used to reduce the dimension of diesel engine operating data of WHTC (the world harmonized transient cycle) bench test, which can make data visualized in three-dimensional space. Then transient diesel engine NOx prediction model is built based on LSTM, and GWO (grey wolf optimizer) is used to optimize the parameters of LSTM. The results showed that R² (determination coefficients) of the GWO-LSTM is 0.987; In the untrained data set, MAE (mean absolute error), MAPE (mean absolute percentage error) and RMSE (root mean square error) are 18.75 × 10^–6, 3.23% and 20.29 × 10^–6, respectively. The same accuracy index are be compared with PSO-BP and static map. It is proved that the GWO-LSTM model can accurately predict the transient NOx emission of diesel engine, and also has good generalization ability with reliability, which provides a reference for software instead of hardware to control diesel engine emission.