Prediction of CO and NOx Emissions from Automotive Engines Based on Machine Learning Algorithms

Abstract

In this paper, five models including BP neural network, Gaussian process regression, support vector machine regression, extreme learning machine and least squares support vector machine are used to model and predict the emissions of carbon monoxide and nitrogen oxides from automobile engines, and the performance of each model is compared. The input data set of this study includes nine parameters (ambient temperature, ambient pressure, ambient humidity, air filter pressure difference, gas turbine exhaust pressure, turbine inlet temperature, turbine outlet temperature, turbine energy yield, and compressor exhaust pressure), and the output data set includes two parameters (carbon monoxide and nitrogen oxides). The comparison of the research results shows that among the five machine learning algorithms, the Gaussian process regression model has the best fitting effect. The model has the highest prediction accuracy and the smallest error. Therefore, the Gaussian process regression model is used to model and predict the emissions of carbon monoxide and nitrogen oxides from automobile engines, and the optimal parameter values of the minimum emissions are found. Applying these automobile engine parameters to daily life is of great significance to alleviating air pollution.