• CN:11-2187/TH
  • ISSN:0577-6686

机械工程学报 ›› 2021, Vol. 57 ›› Issue (20): 161-171.doi: 10.3901/JME.2021.20.161

• 运载工程 • 上一篇    下一篇

扫码分享

功率需求驱动的电动载运装备用动力电池充放电能力预测方法

熊瑞, 闫良基, 王榘   

  1. 北京理工大学机械与车辆学院 北京 100081
  • 收稿日期:2021-01-12 修回日期:2021-05-07 出版日期:2021-10-20 发布日期:2021-12-15
  • 通讯作者: 熊瑞(通信作者),男,1985年出生,博士,教授,IET Fellow,博士研究生导师。主要研究方向为电动载运装备和储能系统。E-mail:rxiong@bit.edu.cn
  • 作者简介:闫良基,男,1996年出生。主要研究方向为动力电池功率能力预测。E-mail:minnowyan@163.com;王榘,男,1991年出生,博士研究生。主要研究方向为新能源汽车动力电池系统管理。E-mail:wang_ju@bit.edu.cn
  • 基金资助:
    国家自然科学基金优秀青年基金资助项目(51922006)。

Power Demand-driven Battery Charging and Discharging Capability Prediction Method for Electric Vehicles

XIONG Rui, YAN Liangji, WANG Ju   

  1. School of Mechanical Engineering, Beijing Institute of Technology, Beijing 100081
  • Received:2021-01-12 Revised:2021-05-07 Online:2021-10-20 Published:2021-12-15

摘要: 充放电功率能力的准确评估是动力电池及电动载运装备安全、高效运行的基础。针对电动载运装备,建立以输入/出功率为控制目标的动力电池模型,描述功率需求驱动的动力电池充放电行为;通过动态优化电池截止电压,提出多步功率预测方法,建立动力电池恒功率需求时的充放电能力预测策略;考虑荷电状态、温度、持续时间等的影响,采用长短期记忆神经网络建立功率修正模型,提升了动力电池充放电功率能力预测性能。结果表明,多步功率预测法能兼顾预测精度和计算效率,最大误差小于3%;全电量区间和宽温度范围内应用效果表明:采用功率修正的功率预测最大误差小于3%,均方根误差低于1%。

关键词: 电动载运装备, 动力电池, 充放电能力, 多步预测法, 长短期记忆神经网络, 功率修正

Abstract: The accurate evaluation of charging and discharging power capability is the basis of safe and efficient operation of the batteries and electric vehicles. Aims at electric transport equipment, the main works are as follow. A battery model with input/output power as the control target is established, and the charging and discharging behaviour of battery-driven by power demand is described. A multi-step power prediction method has been proposed through setting a fixed charge-discharge cut-off control voltage to a dynamic control objective, and the detailed prediction strategy for the charging and discharging power capacity has been established. Considering the influence of the state of charge, temperature, and duration, etc, the power update model is established with the long- and short-term memory neural network to improve the prediction performance of battery charge and discharge power capability. The results show that the proposed method can take into account the prediction accuracy and calculation efficiency, and the maximum error is less than 3%; the power correction method can reasonably predict the power capacity under the full state of charge range, wide temperature, and long duration. The error is less than 3%, and the root mean square error is less than 1%.

Key words: electric vehicles, battery, charge and discharge capacity, multi-step prediction method, long and short-term memory neural network, power correction

中图分类号: