摘要
锂离子电池的容量是评估电池工作状态及能力的重要指标之一,电池容量的精准预测有利于避免电池在非健康状态下工作,延长电池的使用寿命.为了实现锂离子电池容量的快速估计,提出了基于神经网络算法的电池容量快速预测模型.以商用磷酸铁锂电池为研究对象,通过研究循环老化过程中电池充放电数据,提取与电池容量衰减强关联的特征参数作为神经网络输入,结果显示,经200次数据训练后,预测误差小于0.5％.该方法检测时间短,预测精度高,且对连续历史数据依赖度低,易于在能量管理系统(energy management system,EMS)上实现.预测模型中输入的特征参数可以直接由本地数据处理模块计算得到后传输给上位机或云平台,大大减小系统数据传输的压力,提高系统效率.
Abstract
The battery capacity is a key parameter for evaluating the state of health of lithium-ion battery.Accurate capacity estimation can prevent the battery system from working in an unhealthy state,which is crucial for extending the battery life.In this work,a quick capacity estimation approach is proposed by using a model based on an artificial neural network.Two commercial LiFePO4(LFP)battery cells are taken as the object of study here.The features derived from the charging and discharging voltage curves are used as inputs of a three-layer back propagation neural network.The results show that after training by 200 cycles data,a root-mean-square error(RMSE)of less than 0.5％can be achieved.The proposed capacity estimation method can obtain high prediction accuracy with a short detection time cost.Moreover,the algorithm weakly relies on continuous historical data,making it suitable for EMS.The features in the model can be directly calculated by the local data processing module and then transmitted to the upper computer or cloud platform,which greatly reduces the pressure of data transmission and improves the efficiency of the system.
NETL
NSTL
万方数据