基于互信息选择特征向量的锂离池SOH估计
|
摘要
针对锂电池样本呈现出的数据量少、非线性特征,提出了一种基于互信息选择支持向量机回归(SVR)模型的输入特征向量来估计SOH的方法。考虑到影响支持向量机计算结果的因素包含输入样本的代表性和模型参数设置的好坏,在输入样本的选择上使用了互信息的方法,最终选择了恒流恒压充电过程中的电压均值和最高最低温差作为输入特征向量;选择网格搜索算法优化模型参数。实验结果表明,基于互信息选择SVR输入特征向量的锂离池SOH估计结果与基于B P神经网络模型的估计结果相比,所提方法获得了较高的SOH估计精度和泛化能力。
Aiming at the small amount of data and nonlinear characteristics of the samples of lithium battery, a method based on mutual information(MI) that chooses SVR model's input feature vectors is proposed. The error of SVR model's output is influenced by two factors, including the presentation of input samples and the model's parameters. Taking into account these two factors, MI method is determined to be used at input samples' choice. Finally, the mean voltage and maximum and minimum temperature differences in the process of constant current and constant voltage charging are selected as the input feature vector of the model. In addition, the grid search algorithm is selected to optimize the model parameters. Experimental results show that the estimation accuracy and generalization ability of SOH estimation of lithium battery based on SVR of input feature vectors selected by mutual information is better than the BP neural network model.
Aiming at the small amount of data and nonlinear characteristics of the samples of lithium battery, a method based on mutual information(MI) that chooses SVR model's input feature vectors is proposed. The error of SVR model's output is influenced by two factors, including the presentation of input samples and the model's parameters. Taking into account these two factors, MI method is determined to be used at input samples' choice. Finally, the mean voltage and maximum and minimum temperature differences in the process of constant current and constant voltage charging are selected as the input feature vector of the model. In addition, the grid search algorithm is selected to optimize the model parameters. Experimental results show that the estimation accuracy and generalization ability of SOH estimation of lithium battery based on SVR of input feature vectors selected by mutual information is better than the BP neural network model.
引文
[1]刘大同,周建宝,郭力萌,彭宇.锂离子电池健康评估和寿命预测综述[J].仪器仪表学报,2015,01:1-16.Liu D T,Zhou J B,Guo L M,Peng Y.Survey on lithium-ion battery health assessment and cycle life estimation[J].Journal of instrumentand meter,2015,01:1-16.
[2]康燕琼.纯电动汽车锂电池组健康状态(SOH)的估计研究[D].北京交通大学,2015.Kang Y Q.Research on Estimation for SOH of PEV Li-ion Battery Pack[D].Beijing Jiaotong University,2015.
[3] A Eddahech,O Briat, N Bertrand, JY Delétage, JM Vinassa.Behavior and state-of-health monitoring of Li-ion batteries using impedance spectroscopy and recurrent neural networks,International Journal of Electrical Power&Energy Systems,2012,42(1):487-494.
[4] MH Hung, CH Lin, LC Lee, CM Wang.State-of-charge and state-of-health estimation for lithium-ion batteries based on dynamicimpedance techniquejournal of Power Sources,2014,268(4):861-873.
[5]张剑楠.锂离子动力电池健康状态估计算法研究[D].吉林大学,2015.Zhang J N.Study on State of Health Estimation of Lithium-ion Battery[D].Jilin University,2015.
[6] D Andre, A Nuhic, T Soczka-Guth, DU Sauer.Comparative study of a structured neural network and an extended Kalman filter for state of health determination of lithium-ion batteries in hybrid electric vehicles,Engineering Applications of Artificial Intelligence,2013,26(3):951-961.
[7]韩丽,戴广剑,李宁.基于GA-Elman神经网络的电池劣化程度预测研究[J].电源技术,2013,02:249-250+309.Han L,Dai G J,Li N.Prediction of Li-ion Battery's Remaining Capacity Based on GA-SVR Algorithm[J].Power technology,2013,02:249-250+309.
[8]刘苏苏,孙立民.支持向量机与RBF神经网络回归性能比较研究[J].计算机工程与设计,2011,12(2):4202-4205.Liu S S,Sun L M.Performance comparison of regression prediction on support vector machine and RBF neural network[J].Computer engineering and design,2011,12(2):4202-4205.
[9]王宁,谢敏,邓佳梁,刘明波,李嘉龙,王一,刘思捷.基于支持向量机回归组合模型的中长期降温负荷预测[J].电力系统保护与控制,2016,22(3):92-97.Wang N,Xie M,Deng J L,Liu M B,Li J L,Wang Y,Liu S M,Mid-long term temperature-lowering load forecasting based on combination of support vector machine and multiple regression[J].Power system protection and control,2016,22(3):92-97.
[10]范雪莉,冯海泓,原猛.基于互信息的主成分分析特征选择算法[J].控制与决策,2013,06:915-919.Fan X L,Feng H H,Yuan M.PCA based on mutual information for feature selection[J].Control and Design,2013,06:915-919.
[11] T. T. Vo, W. Shen and A. Kapoor.Experimental comparison of charging algorithms for a lithium-ion battery, 2012 10th International Power&Energy Conference(IPEC), 2012,41(23):207-212.
[12] H. Chaoui, N. Golbon, I. Hmouz, R. Souissi and S. Tahar.Lyapunov-Based Adaptive State of Charge and State of Health Estimation for Lithium-Ion Batteries, in IEEE Transactions on Industrial Electronics, 2015, 62(3):1610-1618.
[2]康燕琼.纯电动汽车锂电池组健康状态(SOH)的估计研究[D].北京交通大学,2015.Kang Y Q.Research on Estimation for SOH of PEV Li-ion Battery Pack[D].Beijing Jiaotong University,2015.
[3] A Eddahech,O Briat, N Bertrand, JY Delétage, JM Vinassa.Behavior and state-of-health monitoring of Li-ion batteries using impedance spectroscopy and recurrent neural networks,International Journal of Electrical Power&Energy Systems,2012,42(1):487-494.
[4] MH Hung, CH Lin, LC Lee, CM Wang.State-of-charge and state-of-health estimation for lithium-ion batteries based on dynamicimpedance techniquejournal of Power Sources,2014,268(4):861-873.
[5]张剑楠.锂离子动力电池健康状态估计算法研究[D].吉林大学,2015.Zhang J N.Study on State of Health Estimation of Lithium-ion Battery[D].Jilin University,2015.
[6] D Andre, A Nuhic, T Soczka-Guth, DU Sauer.Comparative study of a structured neural network and an extended Kalman filter for state of health determination of lithium-ion batteries in hybrid electric vehicles,Engineering Applications of Artificial Intelligence,2013,26(3):951-961.
[7]韩丽,戴广剑,李宁.基于GA-Elman神经网络的电池劣化程度预测研究[J].电源技术,2013,02:249-250+309.Han L,Dai G J,Li N.Prediction of Li-ion Battery's Remaining Capacity Based on GA-SVR Algorithm[J].Power technology,2013,02:249-250+309.
[8]刘苏苏,孙立民.支持向量机与RBF神经网络回归性能比较研究[J].计算机工程与设计,2011,12(2):4202-4205.Liu S S,Sun L M.Performance comparison of regression prediction on support vector machine and RBF neural network[J].Computer engineering and design,2011,12(2):4202-4205.
[9]王宁,谢敏,邓佳梁,刘明波,李嘉龙,王一,刘思捷.基于支持向量机回归组合模型的中长期降温负荷预测[J].电力系统保护与控制,2016,22(3):92-97.Wang N,Xie M,Deng J L,Liu M B,Li J L,Wang Y,Liu S M,Mid-long term temperature-lowering load forecasting based on combination of support vector machine and multiple regression[J].Power system protection and control,2016,22(3):92-97.
[10]范雪莉,冯海泓,原猛.基于互信息的主成分分析特征选择算法[J].控制与决策,2013,06:915-919.Fan X L,Feng H H,Yuan M.PCA based on mutual information for feature selection[J].Control and Design,2013,06:915-919.
[11] T. T. Vo, W. Shen and A. Kapoor.Experimental comparison of charging algorithms for a lithium-ion battery, 2012 10th International Power&Energy Conference(IPEC), 2012,41(23):207-212.
[12] H. Chaoui, N. Golbon, I. Hmouz, R. Souissi and S. Tahar.Lyapunov-Based Adaptive State of Charge and State of Health Estimation for Lithium-Ion Batteries, in IEEE Transactions on Industrial Electronics, 2015, 62(3):1610-1618.