【作者】 严英；

【导师】 赵华； 谢辉；

【作者基本信息】 天津大学 ， 动力机械及工程， 2012， 博士

【摘要】 降低运行能耗，延长续驶里程，对于发展纯电动公交客车，应对能源和环境问题，具有重要意义。鉴于司机驾驶行为对整车运行能耗有很大影响，本文深入研究驾驶行为对能耗的影响规律，并提出加速踏板的驾驶特性优化算法，对于降低电动公交客车运行能耗，具有重要的理论意义和实用价值。本文根据课题组提出的公交线路全程多尺度综合规划基本思想，提出了一套实现能量规划、速度规划和驾驶规划的实施方法。根据公交客车进出站的能耗特性，重点提出了基于模型预测的加速踏板驾驶策略优化模型，通过对司机驾驶行为的主动管理，有效降低电动公交客车线路运行能耗。研究开发了基于GPRS网络的电动公交客车运行无线远程监控标定系统，发展了电动公交客车主控制器软硬件。主控制器和无线远程监控标定系统可按照标准协议相互通信，既有效采集车辆线路运行的基本信息，又可实现对加速踏板驾驶策略数据区的远程配置。开发了无线远程通信品质控制算法，改善了数据传送的连续性和准确性，为客车线路运行特性分析和控制策略验证试验提供了保障。依托无线远程监控平台，采集了天津市600路电动公交客车的大量运行数据，系统分析了其工况运行和能耗特征。结果表明，客车出站过程所消耗的能量占到线路总能耗的47%以上。由于司机驾驶行为的差异，导致出站能耗差异可达到28.9%，其根本原因在于，出站过程中加速踏板开度分布影响了电机工况点在效率平面的运行路径。另外，进站过程中司机对车速的预测准确性，对滑行能量回收有决定性的影响。通过对进出站过程中加速踏板的合理规划，减少其加速度的离散程度，实现对电机工况轨迹的优化，是降低电动公交客车能耗的有效途径。系统研究了与能耗相关的司机行为特征时域变量（包括踏板开度、车速、加速度和加速度变化率等参数）的分布情况。根据加速踏板开度信号曲线的倒频谱特征，研究建立了用于司机驾驶行为特征辨识的高斯混合模型(GMM)算法，准确度达到93%以上。构建了基于RBF神经网络的虚拟驾驶员模型，搭建了基于dSPACE的硬件在环仿真实时测试环境。研究提出了基于多维线性空间匹配的扭矩预测方法，依据当前已行驶的片段，在司机驾驶特征向量多维空间中匹配出相似的历史向量轨迹，由此给出司机的扭矩需求，其预测精度达到88%以上。以实际车速与目标车速的偏差最小化为优化目标，对加速踏板的驾驶过程进行规划。仿真表明，驾驶规划降低了行车加速度的离散程度，使能耗改善10.5%。实车验证表明，在司机驾驶操作未受影响的前提下，能耗降低了7.2%。更多还原

【Abstract】 The research of decreasing energy consumption and extension of driving range isa key problem to the development of the pure electric buses besides confronting theenergy crisis and the environmental problems. However, due to inappropriate driverbehaviors the practical endurance mileage of the electric buses are significantlyreduced. In this research, the relationship between the difference of driver behaviorand energy consumption was investigated specifically and it was focused to improvethe practical endurance mileage based on the real-time optimization of driving abilityof the acceleration pedal.The energy optimization algorithm of multi-level programming for the electricbuses, comprised of the driving mission energy programming，the bus station speedprogramming and the preview distance acceleration pedal programming respectively,were put forward, and the pedal programming algorithm was focused to improve theenergy consumption.A data acquisition system based on GPRS network was developed to meet theneed of wireless real-time data recording and remote control variables configuration.An vehicle control unit was developed to drive the electic bus which cancommunicate with the data acquisition system in high communication quality underthe assurance of the wireless communication quality strategy.The relationship between the difference of the driver behavior and energyconsumption was studied through analyzing the main driving fragment of the drivingcycles of the route600electric bus in Tianjin. It shows that47%of the energy wasconsumed in the process of out station and the difference reached28.9%between thedrivers. The essential reason of the difference was the deviaton of the motoroperating path in the efficiency plane which was the result of acceleration pedal usedin different way. The distribution of acceleration pedal was useful to experss theaccuracy of vehicle speed prediction and the tendency of deceleration of drivers in theinlet parking process. The more energy was recycled in the inlet parking process, thehigher efficiency of the energy recovery in the whole driving cycle. An effectivemanagement to the motor operating path in the out station process is very helpful to the improvement of the whole energy consumption besides the minimization of theacceleration dispersion.Several key variables describing the characteristics of driver behavior weredefined, and an effective Gaussian Mixture Model (GMM) identification model of thedrivers, with accuracy higher than93%, was established by extracting the cepstrum ofthe pedal signal besides the key variables mentioned before. A hardware-in-the-loopsimulation model was set up based on dSPACE platform which the key part is thevirtual driver model with real characteristics of the driver based on the Radial BasisFunction (RBF) network.A vector set of driving characteristics of an electric bus driver was defined and amultidimensional linear space was used to describe the running track of the drivingstatus. The torque demand was predicted against historical running tracks withaccuracy of over88%in the driving characteristics vector set of the driver which wasidentified before. The pedal programming algorithm was constructed based on thefuture torque demand prediction. A model predictive control strategy was developedto obtain active management of the driver behavior, in order to minimize the deviationbetween the current and target car speed.Simulation results showed that the acceleration dispersion of the electric bus wasreduced and10.5%of the energy was saved. Experimental results showed that thedriver was not significantly influenced by the regulating process, and7.2%of theenergy was saved.更多还原