Modeling the emission characteristics of the hydrogen-enriched natural gas engines by multi-output least-squares support vector regression: Comprehensive statistical and operating analyses

The hydrogen-enriched natural gas engines (HENGEs) have recently found huge popularity. Despite the broad range of applications of the HENGE, their environmentally-associated problems, like CH₄, CO, and NO_x emissions are not known. Hence, the objective of this study is to model the emission characteristics of HENGEs by the multilayer perceptron neural network (MLPNN) and multi-output least squares support vector regression (MLS-SVR) methods. In this regard, HENGEs emissions are simulated as a function of hydrogen/fuel ratio, engine speed, manifold absolute pressure, excess air ratio, and ignition time. Relevancy analysis showed that the excess air ratio is the most influential factor on both methane and NO_x emission, while the carbon monoxide emission mainly governs by the manifold absolute pressure. Statistical analyses indicate that the MLS-SVR implements this multi-input-multi-output (MIMO) problem more accurately than the MLPNN. The leverage method identifies more than 98% of the experimental datasets as valid measurements. The deployed MLS-SVR estimate 3 × 228 experimentally-measured methane, carbon monoxide, and NO_x emissions with the absolute average relative deviation of 3.55%, 3.30%, and 4.22%, respectively.