基于数据驱动的城市配送电动汽车充电行为分析及模型建立

发布时间：2018-03-17 14:26

本文选题：城市交通　切入点：物流电动汽车　出处：《北京交通大学》2017年硕士论文　论文类型：学位论文

【摘要】：在当前全球汽车工业面临能源短缺和环境污染两大问题的双重挑战下,电动汽车凭借其零污染、低噪声等优势成为人们所关注的焦点。物流业作为我国重要的服务产业,成为电动汽车推广的重要领域。在电动汽车及配套的充电设施产业发展的过程中,出现很多新的问题亟待解决。例如,如何在物流配送中心合理地规划建设充电设施?物流企业如何最优地进行车辆充电的时空分配,以避免产生车辆"充电排队拥堵"的现象?如何规划具有充电约束的物流电动汽车最优的配送路径?这些问题背后蕴含的道理都与城市配送物流电动汽车的充电行为紧密相关,因此分析城市配送物流电动汽车的充电行为规律,建立充电行为模型具有十分重要的理论意义和应用价值。目前国内外对充电行为的研究主要是从电力网系统运行的角度来研究电动汽车的充电行为给电网带来的影响,而从电动汽车出行角度分析电动汽车用户的充电行为的研究成果相对较少。此外,上述研究基本上都是建立在模拟数据或者小样本数据的基础上。而本文基于城市配送物流电动汽车实际运行的海量数据,挖掘城市配送物流电动汽车的充电行为规律,并在此基础上建立了充电行为预测模型,更具有实用性和普适性。为研究城市配送中物流电动汽车用户的充电行为规律,本文采集70辆物流电动车在2014年一年的充放电数据,采用数据删除、插值等一系列的数据处理方法对原始数据进行处理,并识别有效充电过程。在此基础上,采用数据挖掘相关分析方法,建立考虑物流电动汽车充电电量状态的充电行为模型。研究结果表明:用户一般在剩余电量为30%-50%之间时为车辆进行充电,车辆开始充电时的剩余电量服从μ=0.48、σ=0.22的正态分布;车辆开始充电时刻主要集中在14:00-16:00之间。通过数据实验表明,该模型具有较高的准确度。为了准确预测充电时长,本文对采集的数据之间的耦合关系进行了详细的分析。分析表明,充电时长与已充电量SOCc之间存在着明显的正线性相关关系。由此,本文建立了关于已充电量SOCc和充电时长的线性方程。但是在对回归方程的误差序列进行分析时发现,误差序列不符合正态分布。针对这一现象,本文引入时间序列的概念对回归模型进行调整,最终建立了基于回归和ARMA模型的组合预测模型。将两个模型的预测结果进行对比表明,采用组合模型进行充电时长预测比单一的回归模型具有更高的精度和准确度,其残差均落在一条水平带之间,服从正态分布。本文所建立的考虑物流电动汽车的充电电量状态的充电行为模型及充电时长预测模型均具有较高的精确性,同时具有较好的实用性,为车辆的充电调度和用户的出行安排提供科学的决策支持。
[Abstract]:Under the dual challenges of energy shortage and environmental pollution in the global automobile industry, electric vehicles have become the focus of attention because of their advantages of zero pollution and low noise. Logistics industry is an important service industry in China. It has become an important field of electric vehicle promotion. In the process of development of electric vehicle and its associated charging facilities, many new problems need to be solved. For example, how to plan and build charging facilities in logistics distribution center? How to optimize the space-time distribution of vehicle charging in order to avoid the phenomenon of "charging queue congestion"? How to plan the optimal distribution path for logistics electric vehicles with charging constraints? The reasons behind these problems are closely related to the charging behavior of urban distribution logistics electric vehicles, so the law of charging behavior of urban distribution logistics electric vehicles is analyzed. It is of great theoretical significance and practical value to establish a charging behavior model. At present, the research on charging behavior at home and abroad is mainly to study the influence of charging behavior of electric vehicles on the power grid from the point of view of power grid system operation. However, there is relatively little research on the charging behavior of electric vehicle users from the perspective of electric vehicle trip. The above research is basically based on the simulation data or small sample data. However, based on the mass data of the actual operation of the urban distribution logistics electric vehicle, this paper excavates the charging behavior of the urban distribution logistics electric vehicle. On this basis, a charging behavior prediction model is established, which is more practical and universal. In order to study the charging behavior of users of logistics electric vehicles in urban distribution, this paper collects the charging and discharging data of 70 logistics electric vehicles in 2014. A series of data processing methods, such as data deletion, interpolation and so on, are used to process the original data and identify the effective charging process. A charging behavior model considering the charging state of the logistics electric vehicle is established. The results show that the user generally charges the vehicle when the remaining charge is between 30% and 50%, and the residual charge is normally distributed from 渭 0.48 to 蟽 0.22 when the vehicle starts charging. The starting charging time of vehicle is mainly between 14: 00-16: 00. The data experiment shows that the model has high accuracy. In this paper, the coupling relationship between the collected data is analyzed in detail. The analysis shows that there is an obvious positive linear correlation between the length of charging time and the amount of SOCc charged. In this paper, the linear equations of charge amount SOCc and charge time length are established. However, when the error series of regression equation is analyzed, it is found that the error series does not conform to the normal distribution. In this paper, the concept of time series is introduced to adjust the regression model. Finally, the combined prediction model based on regression and ARMA model is established. The combined model has higher precision and accuracy than the single regression model, and the residual error of the model falls between a horizontal band. In this paper, the charging behavior model and the prediction model of charging time of logistics electric vehicle have high accuracy and good practicability. It provides scientific decision support for vehicle charging scheduling and user travel arrangement.
【学位授予单位】：北京交通大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：U491.8

【参考文献】