基于神经网络的并联式混合动力客车动力系统研究

发布时间：2019-06-27 19:47

【摘要】：动力总成匹配技术对混合动力汽车的动力系统起着十分关键的作用,是体现汽车企业自主开发能力的重要方面。本文针对一款济南市传统内燃机客车的动力系统,进行基于神经网络的混合动力总成开发,主要包含有以下研究工作：1.典型工况构建实时采集济南市客车行驶的工况数据后,本文基于两阶聚类分析算法及主成分分析法等理论,对部分实测数据进行分析并构建了客车两阶聚类合成工况;本文对同组数据采用传统的单次聚类方法又构建一个工况,并采用散点分布和建模仿真两种方法对两合成工况与真实济南工况进行了对比,结果均表明两阶聚类合成法构建的工况更接近于车辆实际行驶的道路工况；基于两阶聚类理论基础,本文利用Visual Studio平台C#语言开发了一款构建典型工况的软件,并对大量实测数据处理分析后得到了济南市客车的典型工况。2.混合动力总成匹配在ADVISOR软件平台上,本文构建了该传统内燃机客车的结构模型,经试验验证该模型搭建准确；对客车动力总成动力性、经济性和排放性的需求进行研究后,将该车改造为单轴并联式混合动力客车；之后在已建立的济南市典型工况下,对原有客车和基于电辅助控制策略的混合动力客车进行模拟仿真及台架试验,结果均表明所改造混合动力客车在工况跟踪、动力输出、燃油经济和尾气排放等方面均有大幅改善。3.神经网络控制策略研究基于已建立的混合动力客车模型,利用多层次参数扫描法MLPS优化的逻辑门限值控制策略进行离线仿真,所得数据训练GRBF神经网络;建立基于GRBF神经网络的混合动力客车实时控制策略,并实现ADVISOR二次开发,将GRBF神经网络在不同典型工况下与MLPS优化的逻辑门限值控制策略进行仿真对比,结果表明：本文GRBF神经网络控制策略能够有效降低燃油消耗、减少整车尾气排放、控制电池组SOC平稳变化,有效提高了车辆的效率。
[Abstract]:Powertrain matching technology plays a very key role in the power system of hybrid electric vehicle, and it is an important aspect to reflect the independent development ability of automobile enterprises. In this paper, a hybrid powertrain based on neural network is developed for a traditional internal combustion engine bus in Jinan, which mainly includes the following research work: 1. After collecting the working condition data of Jinan bus in real time, based on the theory of two-order clustering analysis algorithm and principal component analysis, some measured data are analyzed and the second-order clustering synthesis condition of bus is constructed in this paper. In this paper, the traditional single clustering method is used to construct another working condition for the same group of data, and the divergence distribution and modeling and simulation methods are used to compare the two synthetic conditions with the real Jinan working conditions. the results show that the working conditions constructed by the two-order clustering method are closer to the actual road conditions of the vehicle. Based on the theory of two-order clustering, this paper uses Visual Studio platform C # language to develop a software to construct typical working conditions, and obtains the typical working conditions of Jinan bus after processing and analyzing a large number of measured data. In this paper, the structure model of the traditional internal combustion engine bus is constructed on the ADVISOR software platform, and the model is verified to be accurate by experiments. After studying the power performance, economy and emission requirements of the bus powertrain, the car is transformed into a single-axle parallel hybrid electric bus. Then, under the typical working conditions of Jinan city, the simulation and bench test of the original bus and the hybrid electric bus based on the electric auxiliary control strategy are carried out. The results show that the modified hybrid electric bus has been greatly improved in the aspects of working condition tracking, power output, fuel economy and exhaust emission. The research of neural network control strategy is based on the established hybrid electric bus model. The logic threshold control strategy optimized by multi-level parameter scanning MLPS is used to simulate offline, and the obtained data train GRBF neural network. The real-time control strategy of hybrid electric bus based on GRBF neural network is established, and the secondary development of ADVISOR is realized. The GRBF neural network is simulated and compared with the logic threshold control strategy optimized by MLPS under different typical working conditions. The results show that the GRBF neural network control strategy can effectively reduce fuel consumption, reduce vehicle exhaust emissions, control the stable change of battery pack SOC, and effectively improve the vehicle efficiency.
【学位授予单位】：山东大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：U469.7

【相似文献】