基于CNN-BiLSTM的锂电池健康状态估计OA
Prediction of Lithium Batteries SOH Based on CNN-BiLSTM
针对锂电池健康状态 SOH(State of Health,SOH)受多因素耦合影响难以直接测量的问题,构建高精度预测模型以捕捉其退化趋势与关键拐点.提出融合卷积神经网络 CNN(Convolutional Neural Network,CNN)与双向长短期记忆网络 BiL-STM(Bidirectional Long Short-Term Memory,BiLSTM)的 CNN-BiLSTM 模型,利用 CNN 提取电压、电流、温度、阻抗等运行数据的局部空间特征,BiLSTM 挖掘双向时序依赖;基于 NASA 艾姆斯研究中心 ARC-FY08 Q4 标准数据集,对比 LSTM(Long s Hort-Term Memory,LSTM)、BiLSTM、CNN、CNN-LSTM 与 CNN-BiLSTM 性能.结果表明,CNN-BiLSTM 在测试集上的预测精度较其他 4 种显著提高,预测曲线与真实 SOH 变化趋势高度一致,关键退化阶段识别精度超过 90%.CNN-BiLSTM 兼具局部特征提取与双向时序建模优势,可有效预测锂电池 SOH 变化,为电池健康管理及寿命评估提供可靠支撑.
The State of Health(SO H)of lithium-ion batteries is influenced by multiple coupled fac-tors,making it difficult to measure directly.To address this issue and accurately capture the degradation trend and key inflection points,this study constructs a high-precision prediction model.We propose a CNN-BiLSTM model that integrates a Convolutional Neural Network(CNN)and a Bidirectional Long Short-Term Memory network(BiLSTM).The CNN is employed to extract local spatial features from op-erational data,including voltage,current,temperature,and impedance,while the BiLSTM captures bidi-rectional temporal dependencies.Based on the ARC-FY08 Q4 standard dataset from the NASA Ames Re-search Center,we compare the performance of five models:LSTM,BiLSTM,CNN,CNN-LSTM,and CNN-BiLSTM.On the test set,the CNN-BiLSTM model exhibits significantly higher prediction accuracy than the other four models;its predicted curve closely matches the true SOH variation trend,and the rec-ognition accuracy for key degradation stages exceeds 90%.By combining the strengths of local feature ex-traction and bidirectional temporal modeling,the CNN-BiLSTM model can effectively predict lithium-ion battery SOH variations,providing reliable support for battery health management and lifespan assess-ment.
高波;周家睦;董子阳;王兵锐;杜瑶
河北水利电力学院 电力工程系,河北省沧州市黄河西路49号 061001||河北省高校交通基础设施数智化应用技术研发中心,河北省沧州市黄河西路49号 061001河北水利电力学院 电力工程系,河北省沧州市黄河西路49号 061001河北水利电力学院 计算机系,河北省沧州市黄河西路49号 061001河北水利电力学院 电力工程系,河北省沧州市黄河西路49号 061001河北水利电力学院 电力工程系,河北省沧州市黄河西路49号 061001
信息技术与安全科学
锂电池数据驱动健康状态(SO H)CNN-BiLSTM
lithium batterydata-drivenSOHCNN-BiLSTM
《河北水利电力学院学报》 2026 (3)
35-41,7
河北省沧州市科技计划项目(23244101042)河北省教育厅科学研究项目资助(ZC2023092)国家级大学生创新创业项目(202510085002)
评论