基于神经网络的混合动力汽车档位决策研究
发布时间：2019-03-16 19:54
【摘要】：混合动力汽车兼顾了传统汽车和纯电动汽车的优点,能够实现节能减排的目的,是当今新能源汽车领域的研究热点。档位决策是汽车自动变速系统中的重要研究内容,研究智能化、自适应的档位决策方法,对于提高混合动力汽车的动力性、经济性和乘坐舒适性等具有重要的意义。汽车行驶在复杂的人-车-路系统中,在不同的驾驶意图和行驶环境下,如何确定出满足人们对车辆性能需求的最佳换档点存在困难。基于此,本文以单轴并联混合动力汽车为研究对象,引入神经网络对包含驾驶员和环境信息的行车样本数据进行学习和泛化,建立车辆状态参数与最佳档位之间的非线性模型,并结合神经网络的不足,开展了如下研究:(1)分析人-车-路之间的关系,考虑驾驶意图、行驶环境、车辆状态对档位决策的影响,制定了本文的档位决策方案。(2)研究了驾驶意图和行驶环境的识别方法。采集车辆运行状态参数,根据加速踏板信号、制动踏板信号等车辆参数利用模糊推理对驾驶员意图进行了识别,同时利用拉格朗日插值法、移动平均数法等对行驶环境进行了识别。(3)以油门开度、车速、加速度、变速箱输入轴转速为控制参数,档位值为输出,建立神经网络模型。确定神经网络结构,包括网络层数、各层节点个数等。为避免神经网络陷入局部极小,导致网络局部收敛,采用遗传算法对神经网络权值和阈值进行了优化。(4)在MATLAB下,基于识别结果,分别对急加速、上坡、颠簸工况建立神经网络模型,并进行档位值预测仿真与分析。仿真结果表明,训练好的遗传神经网络模型,在特殊驾驶意图和行驶环境下,能够准确地预测汽车档位值;且经遗传算法优化后的神经网络模型,精度有一定的提高。
[Abstract]:Hybrid electric vehicle (HEV), which takes into account the advantages of both traditional and pure electric vehicles, can achieve the goal of energy saving and emission reduction. It is a hot research topic in the field of new energy vehicles. Gear decision-making is an important research content in automatic transmission system of automobile. The research on intelligent and adaptive gear decision-making method is of great significance for improving the power performance, economy and ride comfort of hybrid electric vehicle. In the complicated man-vehicle-road system, it is difficult to determine the optimal shift point to meet people's demand for vehicle performance under different driving intention and driving environment. Based on this, this paper takes the single-axle parallel hybrid vehicle as the research object, introduces the neural network to study and generalize the driving sample data including driver and environment information, and establishes the nonlinear model between the vehicle state parameters and the optimal gear. Combined with the deficiency of neural network, the following research is carried out: (1) the relationship between man-car-road is analyzed, and the influence of driving intention, driving environment and vehicle state on the decision-making of stalls is considered. (2) the identification method of driving intention and driving environment is studied. According to the acceleration pedal signal, brake pedal signal and other vehicle parameters, fuzzy reasoning is used to identify the driver's intention, and Lagrangian interpolation method is used at the same time. The moving average method is used to identify the driving environment. (3) the neural network model is established by taking throttle opening, speed, acceleration, speed of transmission input shaft as control parameters and shift value as output. Determine the structure of neural network, including the number of network layers, the number of nodes in each layer and so on. In order to avoid the neural network falling into local minimum and lead to local convergence of the network, genetic algorithm is used to optimize the weights and thresholds of the neural network. (4) based on the recognition results, the neural networks are accelerated rapidly and up the slope respectively under MATLAB. The model of neural network is established under bumpy condition, and the prediction simulation and analysis of stalls are carried out. The simulation results show that the trained genetic neural network model can accurately predict the vehicle stall value under the special driving intention and driving environment, and the precision of the neural network model optimized by genetic algorithm is improved to a certain extent.
【学位授予单位】：合肥工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：U469.7;TP183
[Abstract]:Hybrid electric vehicle (HEV), which takes into account the advantages of both traditional and pure electric vehicles, can achieve the goal of energy saving and emission reduction. It is a hot research topic in the field of new energy vehicles. Gear decision-making is an important research content in automatic transmission system of automobile. The research on intelligent and adaptive gear decision-making method is of great significance for improving the power performance, economy and ride comfort of hybrid electric vehicle. In the complicated man-vehicle-road system, it is difficult to determine the optimal shift point to meet people's demand for vehicle performance under different driving intention and driving environment. Based on this, this paper takes the single-axle parallel hybrid vehicle as the research object, introduces the neural network to study and generalize the driving sample data including driver and environment information, and establishes the nonlinear model between the vehicle state parameters and the optimal gear. Combined with the deficiency of neural network, the following research is carried out: (1) the relationship between man-car-road is analyzed, and the influence of driving intention, driving environment and vehicle state on the decision-making of stalls is considered. (2) the identification method of driving intention and driving environment is studied. According to the acceleration pedal signal, brake pedal signal and other vehicle parameters, fuzzy reasoning is used to identify the driver's intention, and Lagrangian interpolation method is used at the same time. The moving average method is used to identify the driving environment. (3) the neural network model is established by taking throttle opening, speed, acceleration, speed of transmission input shaft as control parameters and shift value as output. Determine the structure of neural network, including the number of network layers, the number of nodes in each layer and so on. In order to avoid the neural network falling into local minimum and lead to local convergence of the network, genetic algorithm is used to optimize the weights and thresholds of the neural network. (4) based on the recognition results, the neural networks are accelerated rapidly and up the slope respectively under MATLAB. The model of neural network is established under bumpy condition, and the prediction simulation and analysis of stalls are carried out. The simulation results show that the trained genetic neural network model can accurately predict the vehicle stall value under the special driving intention and driving environment, and the precision of the neural network model optimized by genetic algorithm is improved to a certain extent.
【学位授予单位】：合肥工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：U469.7;TP183
【参考文献】
相关期刊论文 前9条
1 周利辉;成艾国;陈涛;赵敏;周泽;;基于BP神经网络的侧碰多目标优化设计[J];中国机械工程;2012年17期
2 李大伟;方锡邦;吴哲;;微型电动汽车AMT动力性换挡规律的制定与仿真分析[J];重庆理工大学学报(自然科学版);2011年01期
3 陈伟根;潘，
本文编号：2441913
本文链接：https://www.wllwen.com/kejilunwen/zidonghuakongzhilunwen/2441913.html
