为了对柴油机的经济性和排放参数进行高效、准确的预测,根据4190型船用柴油机实验数据与边界参数,建立AVL-BOOST甲醇/柴油混合燃料柴油机仿真模型;利用模型进行仿真实验,并建立甲醇掺混比、废气再循环(exhaust gas recirculation,EGR)率、喷油提前角和进气压力4个控制参数对有效油耗率和NOx排放预测数据集;利用该数据集对5种不同核函数的高斯过程回归(Gaussian process regression,GPR)模型进行训练;最后将最优的平方指数高斯过程回归(squared exponential-Gaussian process regression,SE-GPR)模型、AVL-BOOST仿真数据和柴油机实验数据进行对比.结果表明:在数据量为180组时,SE-GPR模型对有效油耗率和NOx排放均取得拟合关联度99％以上,均方根误差(root mean square error,RMSE)分别为 1.859,0.344 5,平均绝对误差(mean absolute error,MAE)分别为 0.954,0.248 9;并且,相较于AVL-BOOST仿真实验,SE-GPR模型对实验数据具有更好的拟合性.
Performance Prediction of Methanol/Diesel Blended Diesel Engine Based on SE-GPR Model
In order to efficiently and accurately predict diesel engine economy and emission parameters,based on the experimental data of the 4190 type marine diesel engine and boundary parameters,an AVL-BOOST simulation model for diesel engines utilizing methanol/diesel blended fuels was established,and a data-set for predicting effective fuel consumption and NOx emissions was created by using this model,incorporating four operational parameters:methanol blending ratio,exhaust gas recirculation(EGR)rate,injection advance angle,and intake pressure.The dataset was employed to train Gaussian process regression(GPR)models with five different kernel functions.Finally,the best-performing squared exponential Gaussian process regression(SE-GPR)model was compared with AVL-BOOST simulation data and diesel engine experimental data.The results showed that the SE-GPR model achieves a correlation of over 99％for both effective fuel consumption and NOx emissions when the dataset contains 180 data sets,with root mean square error(RMSE)values of 1.859,0.344 5,and mean absolute error(MAE)values of 0.954,0.248 9.Moreover,compared to AVL-BOOST simulation experiments,the SE-GPR model exhibits a better fit to the experimental data.
marine diesel enginemethanolGaussian process regressionsquared exponential kernelper-formance prediction
国家科技期刊平台
NSTL
万方数据
