基于IAOA优化XGBoost的柴油机性能预测研究  

Diesel Engine Performance Prediction Based on IAOA-XGBoost Algorithm

作　　者：赵友 王贵勇[2] 申立中[2] 李进龙 曾柏舜 谢亚辉 ZHAO You;WANG Guiyong;SHEN Lizhong;LI Jinlong;ZENG Baishun;XIE Yahui(School of Automotive Engineering,Liuzhou Polytechnic University,Liuzhou 545006,Guangxi,China;Yunnan Key Laboratory of Internal Combustion Engine,Kunming University of Science and Technology,Kunming 650500,China)

机构地区：[1]柳州职业技术大学汽车工程学院,广西柳州545006 [2]昆明理工大学云南省内燃机重点实验室,云南昆明650500

出　　处：《昆明理工大学学报（自然科学版）》2024年第5期108-118,共11页Journal of Kunming University of Science and Technology(Natural Science)

基　　金：国家自然科学基金项目(52066008);云南省科技厅创新引导与科技型企业培育计划专项项目(202104BN050007);广西高校中青年教师科研基础能力提升项目(2022KY1038).

摘　　要：柴油机是一个非线性、高耦合的复杂系统,为了准确预测其性能和变化规律,提出了一种基于改进算术优化算法与极限梯度树结合的性能预测方法.针对算术优化算法本身的缺陷,将莱维飞行、高斯变异和贪心策略融入算法中,提升算法的寻优能力;基于改进后的算术优化算法优化极限梯度树模型的超参数,提升模型的预测精度,形成了一种行之有效的柴油机性能预测方法.研究结果表明:相较于BP神经网络、支持向量机和未优化的极限梯度树模型,经过改进算术优化算法优化的极限梯度树模型有着更高的预测精度,对柴油机比油耗、HC比排放、CO比排放、NO x比排放和涡前排温的预测结果决定系数均大于0.97,且预测值与试验值有较好的相关性.The diesel engine is a highly coupled and nonlinear complex system.To accurately predict its performance and variation,a performance prediction method based on an improved arithmetic optimization algorithm and extreme gradient boosting is proposed.Given the shortcomings of the arithmetic optimization algorithm itself,it is proposed to integrate Levy flight,Gaussian mutation,and greedy strategy into the algorithm to improve its optimization ability.Based on the improved arithmetic optimization algorithm,the hyperparameters of the extreme gradient boosting model are optimized to improve the prediction accuracy of the model,and an effective diesel engine performance prediction method is formed.The results show that:Compared with the BP neural network,support vector machine,and unoptimized extreme gradient boosting model,the extreme gradient boosting optimized by improved arithmetic optimization algorithm has higher prediction accuracy,and the prediction results for specific fuel consumption,specific HC,specific CO,specific NO x and turbine front temperature show that all determination coefficients are greater than 0.97,and the predicted values have a good correlation with the experimental values.

关 键 词：柴油机 极限梯度树 算术优化算法 性能预测 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] TK421[自动化与计算机技术—控制科学与工程]