模型与数据融合的大尺寸快充锂电池分布式温度估计

Estimation of Distributed Temperature of Large-format Fast-charging Lithium-ion Batteries Based on a Model-data Fusion Method

充电时间长、续航里程短、热安全性差等问题正成为阻碍电动汽车大规模应用的主要因素。高安全无损快充和高比能电池技术正成为发展趋势,具体表现为充电功率快速和电池单体大型化提高。然而,大尺寸锂离子电池的温度不一致性问题明显,高功率充电容易导致电池温度快速升高甚至引发热失控。因此,针对快充场景,开发准确高效的温度估计方法尤为重要。针对软包型锂离子电池的分布式温度估计问题,提出一种基于长短期记忆神经网络(Longshort-termmemory,LSTM)与产热模型融合的方法,并在5~40℃的宽温度区间以及多种快充场景下验证了方法的适用性。该方法仅基于一个温度测点的温度信息即可准确估计电池平面其余多个关键测点的温度,可近似获取大尺寸电池二维的温度分布情况,包括最高温度和最大温差,还能有效降低电池成组时的传感器布置成本。融合模型将产热模型结果作为LSTM模型输入,讨论不同温度测点以及不同产热模型作为输入对模型精度的影响。结果表明,温度测点的选择对模型精度影响明显,相比于正极极耳温度,将正极极耳与电池本体连接处温度作为输入,其他测点的估计方均根误差可降低50%,最大误差仅为0.239℃。相较于无产热输入,选择接近电池分布式产热情况的产热模型,估计值方均根误差可减小约11%。

Problems such as long charging time,short mileage,and poor thermal safety are becoming the major factors hindering the large-scale penetration of electric vehicles.High-specific-energy battery technologies and high-safety and damage-free fast charging are becoming popular,reflecting in the larger-sized battery cell and rapidly increased charging power.However,the temperature unevenness within the large-format lithium-ion battery is evident and high-power charging is prone to cause a rapidly elevated temperature and even thermal runaways.Therefore,it is of great importance to propose an accurate and efficient temperature estimation method for fast-charging batteries.A fusion method integrating a long short-term memory(LSTM)network and a heat generation model is proposed for estimating the distributed temperature of pouch-type lithium-ion batteries.The generalizability of the method is verified in a wide temperature ranging from 5 to 40°C and a variety of fast charging scenarios.This method accurately estimates the temperatures of remaining key measurement locations on the battery plane based on the temperature information of only one measurement location.It can not only obtain the approximate two-dimensional temperature distribution of large-format batteries,including the highest temperature and maximum temperature difference,but greatly reduce the cost of temperature sensors when battery grouping.The fusion model takes the output of the heat generation model as the input of the LSTM model,and the effects of the measurement location and heat generation model on the fusion model’s accuracy are discussed.The results show that the selection of temperature measurement locations has a great impact on the prediction accuracy of the model.Compared to the positive tab temperature,taking the temperature measurement at the connection of the positive tab and main battery body as the input can reduce the root mean square error(RMSE)of the remaining measurement locations by 50%,and the maximum RMSE is only 0.239°C.Compared to the model without heat generation as its input,the heat generation model optimally simulating distributed heat generation of the battery can reduce the RMSE of the model by about 11%.