基于LSTM神经网络的混合燃料HCCI发动机复杂工况下燃烧正时估计

Combustion timing estimation of the HCCI engine with mixed fuel under complex operating conditions based on LSTM neural network

均质充气压缩点燃(HCCI)发动机可以通过喷射混合燃料来拓宽工作范围,但也增加了复杂工况下燃烧正时估计的难度。为此,基于长短期记忆(LSTM)神经网络,建立了黑箱模型,用以复杂工况下使用混合燃料的HCCI发动机燃烧正时估计。首先,对多输入数据变量进行Z-Score标准化,使输入数据变换到同一数量级。其次,通过主成分分析法(PCA)对多输入数据变量进行筛选,降低数据维度。最后,用LSTM神经网络对选出的燃烧正时影响相关因素和实际燃烧正时之间的非线性关系进行建模。FTP工况下与其他方法对比验证表明,LSTM模型的各项评价指标均优于SVR和BP模型,其决定系数R2达到了0.98978。

By injecting the mixed fuel into the homogeneous charge compression ignition(HCCI)engine,its working range can be extended.However,the difficulty of combustion timing estimation under complex operating conditions is increased.Therefore,a black box model is formulated,which is based on the long short-term memory(LSTM)neural network.The combustion timing of HCCI engine can be estimated under complex conditions using mixed fuel.Firstly,Z-SCORE standardization is implemented on multi-input data variables.The input data are transformed into the same order of magnitude.Secondly,the principal component analysis is used to filter multi-input data variables to reduce data dimension.Finally,LSTM neural network is utilized to model the nonlinear relationship between the selected combustion timing influencing factors and the actual combustion timing.Compared with other methods under the FTP working condition,results show that the evaluation indicators of the LSTM model are better than those of SVR and BP models.The determination coefficient R^2 reaches 0.98978.