Abstract: Battery state of health (SOH) is an important parameter to characterize battery performance, and accurate SOH estimation is important for battery management and maintenance. The aim of this study is to build a battery SOH estimation model using a long-short-term memory (LSTM) neural network, and to obtain the best model accuracy under different iteration numbers. In this paper, real-time battery operation data are first collected and cleaned and filtered. Then, constant-current charging time, constant-voltage charging time and average discharge voltage are selected as feature indicators to predict the battery health state. By comparing and analyzing the real and predicted values of the three batteries, and the values of mean absolute percentage error (MAPE), root mean square error (RMSE), mean absolute error (MAE) and relative error (RE) evaluation indexes, the accuracy of the three battery models is obtained to be above 98%. The experimental results show that the SOH estimation algorithm based on LSTM has accuracy and feasibility.

Key words: lithium-ion battery; Spearman rank correlation coefficient; battery health status; LSTM neural network

摘要： 电池健康状态（SOH）是表征电池性能的重要参数，准确的 SOH 估计对电池管理和维 护具有重要意义。文章旨在采用长短时记忆模型（LSTM）神经网络搭建电池 SOH 估计模型， 在不同迭代次数条件下得到最佳模型精度。文章首先收集电池实时运行数据并进行清洗和过 滤。然后，选择恒流充电时间、恒压充电时间和平均放电电压等作为特征指标，以预测电池 健康状态。通过对比分析三个电池的真实值与预测值，及平均绝对百分比误差（MAPE）、均 方根差（RMSE）、平均绝对误差（MAE）和相对误差（RE）评价指标的数值，得到三个电池 模型精度均在 98%以上。实验结果表明，基于 LSTM 的 SOH 估计算法具备准确性和可行性。

关键词: 锂离子电池；Spearman 秩相关系数；电池健康状态；LSTM 神经网络