首页|期刊导航|电气技术|基于图形化阻抗特征的锂离子电池健康状态预测方法

基于图形化阻抗特征的锂离子电池健康状态预测方法OA

A graphical impedance feature-based method for state of health prediction of lithium-ion batteries

中文摘要英文摘要

锂离子电池(LIB)是新能源领域关键储能单元,其健康状态(SOH)精准预测对保障系统安全与可靠性具有重要意义.针对现有基于电化学阻抗谱(EIS)的 SOH 预测方法依赖人工特征提取、高维频谱信息挖掘不充分、一维数据特征表达能力有限等问题,本文提出一种基于图形化阻抗特征的锂离子电池 SOH 预测方法.通过格拉姆角场(GAF)将放电电压与 EIS 映射为二维图像,利用卷积神经网络(CNN)提取图像高维空间退化特征、长短期记忆(LSTM)网络刻画时间演化关系,构建 CNN-LSTM 模型,实现 SOH 精准预测.实验结果表明,所提 CNN-LSTM 模型预测结果的平均绝对误差(MAE)为 0.64%,方均根误差(RMSE)为 0.65%,显著优于传统CNN、LSTM 等方法.

As a critical energy storage unit in the new energy sector,accurate state of health(SOH)prediction for the lithium-ion battery(LIB)is paramount to ensure the safety and reliability of battery systems.To address the limitations of existing electrochemical impedance spectroscopy(EIS)-based SOH prediction methods,including heavy reliance on manual feature extraction,insufficient mining of high-dimensional spectral information,and limited feature representation capability of 1 dimensional data,this paper proposes a novel graphical impedance feature-based LIB SOH prediction method.Battery discharge voltage curves and EIS data are mapped into 2 dimensinoal images via Gramian angular field(GAF)to strengthen structured feature representation.A hybrid network is developed,where convolutional neural network(CNN)is adopted to extract high-dimensional spatial degradation features from images and long short term memory(LSTM)is utilized to capture the time-evolution characteristics,thereby realizing accurate SOH prediction of LIBs.The experimental results show that the proposed CNN-LSTM model achieves a mean absolute error(MAE)of 0.64%and a root mean square error(RMSE)of 0.65%,significantly outperforming traditional methods such as CNN and LSTM.

徐超;骆可;张轩铭;孟锦豪

中广核(湖北)新能源投资有限公司,武汉 430062特变电工新疆新能源股份有限公司,西安 710100西安交通大学电气工程学院,西安 710049西安交通大学电气工程学院,西安 710049

锂离子电池(LIB)健康状态(SOH)预测格拉姆角场(GAF)电化学阻抗谱(EIS)

lithium-ion battery(LIB)state of health(SOH)predictionGramian angular field(GAF)electrochemical impedance spectroscopy(EIS)

《电气技术》 2026 (10)

21-27,7

评论