摘要
文章针对锂离子电池健康状态(SOH)估计精度不足与模型泛化能力较弱的问题,提出一种融合Transformer网络与FA-BP优化算法的SOH估测系统。该系统以Transformer结构为核心建模框架,利用其在时间序列特征提取中的优势,提升模型对电池退化趋势的学习能力。引入改进的萤火虫算法与反向传播算法相结合的优化策略,实现模型参数的全局搜索与局部精细调节,提高训练效率与预测稳定性。该方法在SOH估计任务中表现出优越的拟合能力与良好的工程应用潜力,为智能电池管理系统的研究提供了新的技术路径与参考框架。
This paper addresses the issues of insufficient accuracy in estimating the state of health(SOH)of lithium-ion batteries and weak model generalization capabilities.It proposes an SOH estimation system that integrates Transformer networks with FA-BP optimization algorithms.The system uses the Transformer architecture as the core modeling framework and leverages its advantages in time series feature extraction to enhance the model's learning ability regarding battery degradation trends.An improved optimization strategy that combines Firefly Algorithm and Backpropagation algorithms is introduced,enabling global search and local fine-tuning of model parameters,thereby improving training efficiency and prediction stability.This method demonstrates superior fitting capabilities and good engineering application potential in the SOH estimation task,providing a new technical path and reference framework for the research on intelligent battery management systems.
出处
《时代汽车》
2025年第14期121-123,共3页
Auto Time
基金
大学生创新创业训练计划项目“基于深度学习与FA-BP优化算法的锂离子电池SOH的估测”(项目编号:S202410564039)。
