面向燃烧闭环控制的天然气掺氢发动机CA50预测

Key Parameter CA50 Prediction of Hydrogen-enriched Compressed Natural Gas Engine for Combustion Closed-loop Control

摘要:

为探究提高发动机效率和降低排放的方法，开展了燃烧闭环控制关键参数CA50对天然气掺氢混合燃料（HCNG）发动机燃烧和排放影响的试验研究，并基于试验结果对CA50进行统计分析。利用粒子群优化反向传播神经网络（PSO-BPNN）算法对CA50进行预测，并探究了混合策略优化对PSO-BPNN模型预测性能的影响。结果表明，CA50对HCNG发动机的燃烧特性和排放有显著影响；CA50服从正态分布，不存在自相关，可作为燃烧闭环控制的反馈参数；通过PSO-BPNN方法建立的CA50预测模型具有较高的预测性能和良好的泛化能力，平均绝对误差为0.25°CA，相关系数大于0.997；混合策略可在不降低预测精度的情况下显著提高模型的收敛速度，CPU运行时间最多可缩短73.02%。

Abstract:

To explore the method of improving engine efficiency and reducing emissions， the impact of combustion closed-loop control key parameters CA50 on the combustion and emissions of a hydrogen-enriched compressed natural gas （HCNG） engine was experimentally studied， and CA50 based on the experimental results was statistically analyzed. Meanwhile， the particle swarm optimization （PSO） back-propagation neural network （BPNN） algorithm was applied to the prediction of CA50， and the influence of hybrid strategy optimization on the performance of PSO-BPNN model was investigated. Results show that： CA50 has a significant impact on the combustion characteristics and emissions of the HCNG engine； CA50 obeys the normal distribution and has no auto-correlation， so it can be used as the feedback parameter of combustion closed-loop control； the CA50 prediction model established by PSO-BPNN method has the high prediction performance and good generalization ability， with the average absolute error of 0.25°CA and the correlation coefficient of more than 0.997； the hybrid strategy can significantly improve the convergence speed of the model without reducing prediction accuracy， with the CPU running time reduced by up to 73.02%.