多因素影响下融合RNN和AUKF的 矿用锂离子电池SOC估计  

SOC estimation of mine-used lithium-ion batteries by integrating RNN and AUKF under influence of multiple factors

作　　者：窦元运 张成知 封居强 DOU Yuanyun;ZHANG Chengzhi;FENG Juqiang(School of Materials Science and Engineering,Hefei University of Technology,Hefei Anhui 230009,China;School of Mechanical and Electrical Engineering,Huainan Normal University,Huainan Anhui 232008,China)

机构地区：[1]合肥工业大学材料科学与工程学院,安徽合肥230009 [2]淮南师范学院机械与电气工程学院,安徽淮南232008

出　　处：《电源技术》2025年第4期764-771,共8页Chinese Journal of Power Sources

基　　金：安徽省科技重大专项(2021jyxm1368);安徽省高校中青年教师培养行动项目(YQYB2023030);淮南师范学院校级科研项目(2024XJZD012)。

摘　　要：针对矿用锂离子电池在实际应用中面临的荷电状态(SOC)估计难题,提出了一种结合递归神经网络(RNN)和自适应无迹卡尔曼滤波(AUKF)的新方法,该方法考虑了温度、倍率等多因素对SOC估计的影响。对228 Ah大容量矿用锂离子电池进行多因素影响实验,构建改进的一阶RC等效电路模型。利用RNN回归分析多因素对OCV-SOC关系及模型参数的影响。采用AUKF算法对电池在不同复杂工况下的模型进行有效辨识和SOC估计。实验结果表明,该方法能够显著提高矿用锂离子电池SOC估计的准确性和鲁棒性。研究结果可为矿用设备的智能化管理和维护提供重要的技术支持。Addressing the challenge of state of charge(SOC)estimation faced by lithium-ion batteries in mining applications,this paper proposed a novel approach that combines recurrent neural networks(RNN)with adaptive unscented Kalman filtering(AUKF).This method takes into account the influence of multiple factors,such as temperature and rate,on SOC estimation.Multi-factor impact tests were conducted on a 228 Ah high-capacity lithium-ion battery for mining applications,leading to the development of an improved first-order RC equivalent circuit model.RNN regression was utilized to analyze the impact of multiple factors on the SOC-open circuit voltage(OCV)relationship and model parameters.The AUKF algorithm was employed to effectively identify the battery model and estimate the SOC under various complex operating conditions.The experimental results demonstrate that this method significantly improves the accuracy and robustness of SOC estimation for lithium-ion batteries in mining applications.The findings of this paper can provide crucial technical support for intelligent management and maintenance of mining equipment.

关 键 词：SOC估计 矿用锂离子电池 多因素 递归神经网络 自适应无迹卡尔曼滤波 

分 类 号：TM912[电气工程—电力电子与电力传动]