采用机器学习的天然气掺氢发动机可预测燃烧模型

打开文本图片集

中图分类号：TK432 文献标志码：A

DOI：10.7652/xjtuxb202601015 文章编号： 0253-987X（2026）01-0150-12

Predictive Combustion Model for Hydrogen-Enriched Compressed Natural Gas Engines Using Machine Learning

LI Hongzhi1，HU Zhenyul²，TAO Weijun1，HUANG Yongchengl （1. School of Energy and Power Engineering，Xi'an Jiaotong University，Xi'an 71o049，China; 2.BYD Auto Industry Co.，Ltd.，Shenzhen，Guangdong ，China）

Abstract： To address the lack of fundamental combustion data for hydrogen-enriched compressed natural gas（HCNG） in engine one-dimensional performance simulation software GT-Power，and to accurately simulate the effects of HCNG’s unique physicochemical properties on engine performance，a novel machine learning-driven predictive combustion model is developed. Firstly， a computational model for the laminar burning velocity and ignition delay time of HCNG is developed based on a machine learning approach utilizing an artificial neural network （ANN）. Subsequently， the model undergoes error analysis and validation，and is implemented as subroutines coupled with GT-Power software. Finally，the newly developed model is calibrated and validated against experimental data. Using this model， the influence of hydrogen volume fraction φ（H2 ）（hydrogen blending ratio） on engine performance is investigated under external characteristic conditions. The results indicate that： The correlation coefficients between the laminar burning velocity and ignition delay time predicted by the trained ANN combustion model and the original values both exceed O.99；the error in engine performance parameters calculated using the calibrated model is less than 3% ，with errors in combustion phasing and knock-limited spark advance angle being less than 2∘ ;limited by knock and pre-turbine exhaust temperature， the maximum allowable φ（H2 ）is 20% ；As φ（H2） ）increases，the optimal spark timing is gradually delayed，and the effective thermal eficiency decreases; hydrogen enrichment accelerates combustion but increases knock tendency， while Exhaust Gas Recirculation （EGR） slows combustion but effectively suppresses knock and advanced combustion phasing，with the synergistic control of both hydrogen blending and EGR improving the engine's effective thermal efficiency.

Keywords： predictive combustion model; hydrogen-enriched compressed natural gas; artificial neural network；laminar burning velocity；ignition delay time

近年来，为顺利实现“双碳”目标，在《2023年前碳达峰行动方案》的指导下，我国大力推动天然气和氢气等清洁能源在交通运输领域的发展[1]。（剩余17952字）