基于特征工程与改进随机森林回归的锂离子电池SOH预测

doi:10.19595/j.cnki.1000-6753.tces.251060

Abstract
Figure/Table
References
Related Citation (15)

Download: PDF (0 KB) HTML (1 KB)
Export: BibTeX | EndNote (RIS)

Abstract To address the limitations of insufficient feature characterization and hyperparameter sensitivity in existing state-of-health (SOH) prediction methods for lithium-ion batteries, this paper proposes a novel SOH prediction framework integrating a graph attention network (GAT), a conditional generative adversarial network (CGAN), improved beluga whale optimization (IBWO), and random forest regression (RFR).
Firstly, using lithium-ion batteries from the publicly available NASA and MIT battery datasets, voltage, current, temperature, and actual SOH data are collected. Subsequently, the GAT method is employed to adaptively learn aging trends from the voltage, current, and temperature data, extracting health features (HFs) that effectively reflect battery degradation. Then, optimal HF is screened through Pearson correlation coefficient analysis.
Secondly, to address the insufficient nonlinear correlation between the HFextracted by GAT and SOH, this study proposes a feature enhancement framework based on a CGAN. The generator generates feature vectors consistent with the original SOH distribution via adversarial learning, while the discriminator dynamically identifies feature sources using a convolutional architecture. Through iterative adversarial training, the correlation between generated features and degradation patterns progressively approximates the real SOH distribution. Thus, precise tracking of complex battery degradation trajectoriesis achievable.
Then, the traditional beluga whale optimization (BWO) algorithm is improved by introducing three optimization strategies: Circle chaotic mapping, quasi-oppositional learning, and dynamic weight adjustment. To validate the correctness and effectiveness of the IBWO algorithm, unimodal and multimodal functions from the CEC2005 test suite are selected for performance comparisons with IBWO, traditional BWO, the whale optimization algorithm (WOA), and particle swarm optimization (PSO). The results show that the proposed IBWO algorithm achieves superior convergence speed, optimization accuracy, and stability in the search for optimal solutions. Subsequently, the IBWO algorithm is employed to optimize the hyperparameters of RFR, establishing the SOH prediction model.
Finally, module-level comparative experiments demonstrate that the GAT-CGAN-IBWO-RFR model effectively keeps the SOH prediction error (PE) within [-0.005, 0.005], indicating significant error suppression. Increasing the training dataset size improves SOH prediction accuracy, with mean absolute error (MAE), mean absolute percentage error (MAPE), and root mean square error (RMSE) consistently remaining below 0.6%. Furthermore, cross-study comparisons with existing models confirm the GAT-CGAN-IBWO-RFR model's superior performance in SOH prediction. The proposed model's strong generalization capability and high- precision SOH prediction performance are verified.

Key words： Lithium-ion batteries state of health estimation graph attention network conditional generative adversarial network improved beluga whale optimization algorithm random forest regression

Received: 17 June 2025

PACS:

TM912

	Service

	E-mail this article
	Add to my bookshelf
	Add to citation manager
	E-mail Alert
	RSS
	Articles by authors
	Fan Xingming
	Huang Wenxi
	Zhang Xin
	Zhu Gezhi

Cite this article:

Fan Xingming,Huang Wenxi,Zhang Xin等. A Hybrid State of Health Estimation Framework for Lithium-Ion Batteries Integrating Feature Engineering and Optimized Random Forest Regression[J]. Transactions of China Electrotechnical Society, 2026, 41(12): 4286-4297.

URL:

https://dgjsxb.ces-transaction.com/EN/10.19595/j.cnki.1000-6753.tces.251060 OR https://dgjsxb.ces-transaction.com/EN/Y2026/V41/I12/4286

[1] 范兴明, 吴润玮, 封浩, 等. 基于VAUKF的锂离子电池SOC预测研究[J]. 电工技术学报, 2025, 40(6): 1974-1983.
Fan Xingming, Wu Runwei, Feng Hao, et al.Research on SOC prediction of lithium-ion batteries based on VAUKF[J]. Transactions of China Electrotechnical Society, 2025, 40(6): 1974-1983.
[2] 陈媛, 段文献, 何怡刚, 等. 带降噪自编码器和门控递归混合神经网络的电池健康状态估算[J]. 电工技术学报, 2024, 39(24): 7933-7949.
Chen Yuan, Duan Wenxian, He Yigang, et al.State of health estimation of lithium ion battery based on denoising autoencoder-gated recurrent unit[J]. Trans-actions of China Electrotechnical Society, 2024, 39(24): 7933-7949.
[3] 孙玉树, 龚一莼, 董亮, 等. 电池储能系统状态估计综述[J]. 中南大学学报(自然科学版), 2024, 55(6): 2320-2333.
Sun Yushu, Gong Yichun, Dong Liang, et al.Review of state estimation for battery energy storage systems[J]. Journal of Central South University (Science and Technology), 2024, 55(6): 2320-2333.
[4] 金帅, 董静. 基于数据驱动的锂离子电池健康状态估计研究进展综述[J]. 仪器仪表学报, 2024, 45(3): 45-59.
Jin Shuai, Dong Jing.Review on progress of data-driven based health state estimation for lithium-ion batteries[J]. Chinese Journal of Scientific Instrument, 2024, 45(3): 45-59.
[5] 刘萍, 李泽文, 蔡雨思, 等. 基于等效电路模型和数据驱动模型融合的SOC和SOH联合估计方法[J]. 电工技术学报, 2024, 39(10): 3232-3243.
Liu Ping, Li Zewen, Cai Yusi, et al.Joint estimation method of SOC and SOH based on fusion of equivalent circuit model and data-driven model[J]. Transactions of China Electrotechnical Society, 2024, 39(10): 3232-3243.
[6] 赵轩, 李美莹, 余强, 等. 电动汽车动力锂电池状态估计综述[J]. 中国公路学报, 2023, 36(6): 254-283.
Zhao Xuan, Li Meiying, Yu Qiang, et al.State estimation of power lithium batteries for electric vehicles: a review[J]. China Journal of Highway and Transport, 2023, 36(6): 254-283.
[7] 马君, 万俊杰. 基于健康特征筛选与GWO-LSSVM的锂电池健康状态预测[J]. 电气技术, 2024, 25(2): 37-44.
Ma Jun, Wan Junjie.Prediction of state of health for lithium battery based on health feature screening and GWO-LSSVM[J]. Electrical Engineering, 2024, 25(2): 37-44.
[8] 吴伟丽, 卢双双. 基于PCA-CEEMDAN和改进SVR的锂电池RUL预测[J]. 电源学报, 2025, 23(6): 267-280.
Wu Weili, Lu Shuangshuang.RUL prediction of lithium-ion batteries based on PCA-CEEMDAN and improved SVR[J]. Journal of Power Supply, 2025, 23(6): 267-280.
[9] Wang Yuhang, Huang Haihong, Wang Haixin, et al.A fast estimation method for state-of-health of retired batteries based on health features[J]. Journal of Energy Storage, 2023, 72: 108677.
[10] Bharath Y K, Anandu V P, Vinatha U, et al.Long-term estimation of SoH using cascaded LSTM-RNN for lithium batteries subjected to aging and accelerated degradation[J]. Energy Storage, 2024, 6(8): e70066.
[11] Zhang Lijun, Ji Tuo, Yu Shihao, et al.Accurate prediction approach of SOH for lithium-ion batteries based on LSTM method[J]. Batteries, 2023, 9(3): 177.
[12] Sun Shuo, Sun Junzhong, Wang Zongliang, et al.Prediction of battery SOH by CNN-BiLSTM network fused with attention mechanism[J]. Energies, 2022, 15(12): 4428.
[13] 武骥, 方雷超, 刘兴涛, 等. 基于特征优化和随机森林算法的锂离子电池SOH估计[J]. 机械工程学报, 2024, 60(12): 335-343.
Wu Ji, Fang Leichao, Liu Xingtao, et al.State of health estimation of lithium-ion battery based on feature optimization and random forest algorithm[J]. Journal of Mechanical Engineering, 2024, 60(12): 335-343.
[14] 张少宇, 伍春晖, 熊文渊. 采用门控循环神经网络估计锂离子电池健康状态[J]. 红外与激光工程, 2021, 50(2): 20200339.
Zhang Shaoyu, Wu Chunhui, Xiong Wenyuan.State of health estimation for lithium-ion batteries using recurrent neural networks with gated recurrent unit[J]. Infrared and Laser Engineering, 2021, 50(2): 20200339.
[15] 李英顺, 阚宏达, 郭占男, 等. 基于数据预处理和VMD-LSTM-GPR的锂离子电池剩余寿命预测[J]. 电工技术学报, 2024, 39(10): 3244-3258.
Li Yingshun, Kan Hongda, Guo Zhannan, et al.Prediction of remaining useful life of lithium-ion battery based on data preprocessing and VMD-LSTM-GPR[J]. Transactions of China Electrotech-nical Society, 2024, 39(10): 3244-3258.
[16] Sayed K, Gronfula M G, Elsayed M M, et al.hyperparameter optimization for random forest and support vector machine models using particle swarm optimization for battery health monitoring and predictive maintenance[J]. Available at SSRN 5177792.
[17] Severson K A, Attia P M, Jin N, et al.Data-driven prediction of battery cycle life before capacity degradation[J]. Nature Energy, 2019, 4(5): 383-391.
[18] Zhong Changting, Li Gang, Meng Zeng.Beluga whale optimization: a novel nature-inspired metaheuristic algorithm[J]. Knowledge-Based Systems, 2022, 251: 109215.
[19] Zhao Jiahui, Zhu Yong, Zhang Bin, et al.Method of predicting SOH and RUL of lithium-ion battery based on the combination of LSTM and GPR[J]. Susta-inability, 2022, 14(19): 11865.
[20] Yuan Zifan, Tian Tian, Hao Fuchong, et al.A hybrid neural network based on variational mode decom-position denoising for predicting state-of-health of lithium-ion batteries[J]. Journal of Power Sources, 2024, 609: 234697.
[21] Shao Liyuan, Zhang Yong, Zheng Xiujuan, et al.SOH estimation of lithium-ion batteries subject to partly missing data: a Kolmogorov-Arnold-Linformer model[J]. Neurocomputing, 2025, 638: 130181.
[22] Ansari S, Ayob A, Hossain Lipu M S, et al. Jellyfish optimized recurrent neural network for state of health estimation of lithium-ion batteries[J]. Expert Systems with Applications, 2024, 238: 121904.