考虑健康状态指标与“记忆效应”的镉镍蓄电池SOC估算模型

State of charge estimation of Ni-Cd batteries with consideration of health indicator and memory effect

如何通过在线监测的直接测量参数准确估计电池荷电状态(State of Charge,SOC)与健康状态(State of Health,SOH),是蓄电池管理系统建立的核心与关键。设计的SOC估算方法为镉镍蓄电池管理系统有效地监测蓄电池组性能状态和寿命状态提供基础,有助于动车组在运行过程中的安全预警;同时,为蓄电池检修与维护策略优化提供数据支撑,助力国家可持续发展战略。从蓄电池不同寿命阶段内的充电起始电压序列中提取出可描述当前最大容量的潜在特征,通过F检验(F-Test)与主成分分析(Principal Component Analysis,PCA)进行特征筛选与特征融合,获得蓄电池健康状态指标;由充放电循环试验中采集到的不同寿命阶段内的放电终止电压建立镉镍蓄电池“记忆效应”的近似表达函数;基于此,采用基于Bagging的随机森林构建放电过程中蓄电池两端电压与SOC间的关联模型,可在蓄电池放电过程中实现基于放电电压的SOC估算。最终,试验结果显示:通过SOC估算值与实际测量值的对比,得到模型均方根误差(Mean Square Error,MSE)和平均绝对误差百分比(Mean Absolute Percentage Error,MAPE)分别为0.1486和0.8112%,证明了所提出的SOC估算模型取得了较高的估算精度与较强的鲁棒性,为在线监测镉镍蓄电池SOC提供基础。

How to estimate the state of charge(SOC)and state of health(SOH)accurately through the direct measurement parameters of on-line monitoring is the core and key part in battery management system.The SOC estimation method designed in this paper provided a basis for Ni-Cd battery management system to effectively monitor the performance state and lifetime prediction of the battery,which developed security pre-warning mechanism of Electric Multiple Units(EMU).Meanwhile,it could provide data support for battery overhaul and maintenance strategy optimization and contributes to national sustainable development strategy.In this paper,the potential features that could describe the current maximum capacity were extracted from the initial voltage series in different life stages of the battery.The health indicators of the battery were obtained by feature selection and feature fusion through F-test and Principal Component Analysis(PCA).The approximate linear expression function of memory effect of Ni-Cd battery was established by the discharge termination voltage sequence collected in the charge-discharge cycle test.From this,the relevancy model between discharge voltage values and SOCs in the discharge process had been built by random forest based on Bagging.The mean square error(MSE)and mean absolute percentage error(MAPE)of the model for SOC estimation error on test sets were 0.1486 and 0.8112,respectively.This result illustrates the validity of this model,and proves that this method can obtain good performance with high accuracy and robustness,which provides the basis for on-line monitoring SOC of Ni-Cd battery.