并联式混合动力发动机神经网络法转矩预测与闭环控制

Torque Estimation and Closed-Loop Control of Parallel Hybrid Engine Using ANN Method

摘要:

利用实际发动机的标定数据搭建了GT-Suite及Matlab/Simulink联合仿真模型，建立了基于进气和发动机状态参数的预测转矩反馈协同控制模块。对比了ANN法和现有发动机的标定脉谱插值预测（MAP）法2种方法下发动机稳态及瞬态转矩变化、升降挡等工况预测的结果误差，结果表明：稳态工况下MAP法较为可靠，低、中、高3种发动机转速下转矩预测波动小，误差比ANN法低1.31%、1.09%和1.52%；实际瞬态转矩跃变及阶跃工况下，ANN法较MAP法误差低5.62%和1.32%，升降挡工况下低1.93%和0.84%。

Abstract:

In this paper， a joint simulation model of GT-suite and MATLAB / Simulink was constructed by using the calibration data of the actual engine and a collaborative control module for estimated torque feedback based on intake air and engine state parameters was established. A comparison was made between the estimation results errors of the ANN method and the Map method for estimating the steady state， transient torque variation， upshift and downshift of the engine. The results show that the Map method is more reliable under steady-state conditions， and the error of ANN method is small at low， medium， and high engine speeds， with errors of 1.31%， 1.09%， and 1.52% lower than the ANN method， and the error of the ANN method is 5.62% and 1.32% lower than that of the Map method under torque transient conditions， and 1.93% and 0.84% lower than that of the Map method under lifting conditions.