摘要
电动汽车的锂离子电池具有非线性和时变性,加上实际运行中的动态操作和环境影响,导致电池外部参数(电流、电压和温度)的不稳定变化,使得SOC的准确估计变得充满挑战。提出使用卷积神经网络(Convolutional Neural Network,CNN)和双向长短期记忆(Bidirectional Long Short-term Memory,BiLSTM)网络结合的神经网络架构,建立SOC估计模型,通过基于TPE过程的贝叶斯优化找到最优的超参数组合,提升SOC估计的精度。通过使用从动态应力测试(Dynamic Stress Test,DST),联邦城市驾驶(Federal Urban Driving Schedule,FUDS)和US06测试三种工况中收集的数据进行训练和测试,得到SOC估计均方根误差小于1.22%,平均误差小于1.08%,最大误差小于3.74%,R2大于0.9982。结果表明该模型在未知工况和温度下具有很好的泛化性,鲁棒性和估计精度。
The lithium-ion battery of electric vehicles is nonlinear and time-varying.Coupled with dynamic oper⁃ations and environmental impacts in actual operation,it leads to unstable changes in external battery parameters(cur⁃rent,voltage,and temperature),making the accurate estimation of SOC full of challenges.This paper presents a neural network architecture that combines Convolutional Neural Network(CNN)with Bidirectional Long Short-Term Memory(BiLSTM)network to establish a SOC estimation model.Bayesian optimization based on the Tree-structured Parzen Estimator(TPE)process is utilized to find the optimal hyperparameter combination,enhancing the precision of SOC estimation.Data collected from the Dynamic Stress Test(DST),Federal Urban Driving Schedule(FUDS),and US06 test cycles were used to train and test the model,which achieved a root mean square error(RMSE)for SOC estimation of less than 1.22%,an average error below 1.08%,a maximum error under 3.74%,and an R-squared(R2)value greater than 0.9982.The results indicate that the model exhibits excellent generalizability,robustness,and estimation accuracy under unknown operating conditions and temperatures.
作者
刘志强
臧阿顺
LIU Zhi-qiang;ZANG A-shun(Changsha University of Science and Technology,Changsha Hunan 410114,China)
出处
《计算机仿真》
2025年第7期229-235,共7页
Computer Simulation
基金
国家自然科学基金(51976016)。
关键词
锂离子电池
荷电状态估计
贝叶斯优化
卷积神经网络
双向长短期记忆网络
Lithium-ion battery
State of charge estimation
Bayesian optimization
Convolutional neural network
Bidirectional long short-term memory network
