用于控制的发动机转矩估计方法研究  

Methods of Engine Torque Estimation for Control Algorithms

作　　者：杜常清[1] 颜伏伍[1] 严运兵[1] 杨平龙[1] 

机构地区：[1]武汉理工大学汽车工程学院,湖北武汉430070

出　　处：《内燃机学报》2008年第5期446-451,共6页Transactions of Csice

基　　金：国家863计划(2005AA501220-3)

摘　　要：有一定精度和实时性的发动机动态转矩估计模型,对机电混合电动汽车动力及传动系统控制有重要意义。对发动机动态特性的影响因素进行了分析;分析了平均值模型的优缺点,针对混合动力控制和建模仿真的需要,对模型进行了有效的简化,大大提高了模型的实用性;分析了混合动力中发动机的特定动态工况,建立了特定动态工况下的发动机神经网络模型;基于发动机稳态和动态试验数据,对平均值模型和神经网络模型进行了实证研究。结果表明,基于神经网络的简化平均值模型适用于发动机稳态工况和节气门开度变化率不大的动态工况;神经网络模型适用于发动机稳态工况和特定的动态工况。Accuracy and real-time dynamic engine model for torque estimation is very important for the development of hybrid electric vehicle powertrain. The influencing factors of dynamic engine characteristic were analyzed. To meet the requirement of powertrain control in hybrid electric vehicles, the mean value engine model was simplified based on the analysis of the model＇s characteristics, and model applicability was improved. The specific dynamic operating condition of engine in hybrid powertrain system was also analyzed, and an engine model was built by BP neural network to satisfy the corresponding requirement. These two engine models were validated via both steady and dynamic engine testing on a dynamic engine test bench. The study shows that the mean value engine model simplified by neural network is applicable for both steady and dynamic engine operation conditions without fast varying throttle position, and the neural network model is applicable for steady and specified dynamic operating condition.

关 键 词：发动机 转矩估计 平均值模型 BP神经网络 

分 类 号：TK421[动力工程及工程热物理—动力机械及工程]