面向能量管理的燃料电池混合动力模型辨识  

Model Identification Algorithm of Hybrid Fuel Cell System Facing Energy Management Optimization

作　　者：徐梁飞[1] 包磊[1] 华剑锋[1] 李建秋[1] 欧阳明高[1] 

机构地区：[1]清华大学汽车安全与节能国家重点实验室,北京100084

出　　处：《系统仿真学报》2009年第17期5361-5365,共5页Journal of System Simulation

基　　金：国家"八六三"计划"节能与新能源汽车重大专项"(2006AA11A102)

摘　　要：燃料电池混合动力能量管理算法离线及在线优化需要建立一套有效的模型辨识方法。根据混合动力系统的部件特性,使用TTCAN数据,基于最小二乘原理建立系统辨识算法,辨识出系统模型。离线仿真结果表明,所辨识的模型能反映当前系统性能,可作为能量管理算法优化的依据。为提高能量管理算法的自适应调整能力,需要将离线辨识改为在线辨识,并整合到整车控制器中。在Matlab/Simulink环境中搭建在线辨识算法,并通过自动代码生成与整车控制器底层代码无缝结合。增加了在线辨识算法的能量管理算法运行时间从原先的0.58ms提高到3.2ms,但仍然小于TTCAN网络的最小闲置时间4ms。在线辨识算法虽然增加了整车控制器运算负荷,但并未影响TTCAN通讯时序。中国城市公交典型工况测试表明,采用在线辨识算法后的系统氢气消耗从8.2kg/100km降低到7.9kg/100km,约降低3.6%。为控制燃料电池输出功率波动,进一步工作中需要将燃料电池模型加入在线辨识算法,并采用递推算法降低整车控制器计算负荷。In order to optimize the energy management strategy, a model identification algorithm for the hybrid fuel cell system is strongly demanded. According to the components＇ properties, a model identification algorithm based on the least square principle was developed. Off-line simulation results show that, the identified model reflects the system properties well An on-line identification algorithm was further developed in Matlab/Simulink to improve the self-adaptive ability of energy management strategy. Codes were automatically generated and integrated into Vehicle Control Unit （VCU）. With the on-line identification algorithm, the computing time of VCU increased from 0.58ms to 3.2ms, which was still less than the minimal TTCAN idle time- 4ms. Therefore, the on-line algorithm increased the computing burden, but didn＇t influence the TTCAN communication quality. Result of China typical city cycle testing shows that, the fuel economy was improved from 8.2kg/100kin to 7.9kg/100km with the on-line identification algorithm.

关 键 词：燃料电池混合动力系统 模型辨识 最小二乘法 在线辨识 

分 类 号：U469.7[机械工程—车辆工程]