Scholarlink

An in-depth discussion on the accurate estimation method of lithium battery State of Charge (SOC) is carried out, and for SOC estimation, a composite method of BP neural network with dynamic adaptive learning rate and Dynamic Adaptive Learning Rate BP Neural Network Modified Untraceable Kalman Filter (DABP-U) is proposed. The paper first employs the Unscented Kalman Filter (UKF) algorithm to make a preliminary estimation of the battery SOC, and then corrects and optimizes the UKF estimation results using a BP neural network with a dynamic adaptive learning rate, thereby improving the accuracy of the SOC estimation. The simulation and validation results show that the DABP-U method has higher estimation accuracy, stability, and faster convergence speed compared with the traditional algorithm, and especially performs well in dealing with the uncertainty of battery model parameters and the variation of the external environment.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Accurate Estimation of SOC by BP Neural Network Modified UKF Algorithm with Dynamic Adaptive Learning Rate

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Chao Yang,
  • Xin Xiong

摘要

An in-depth discussion on the accurate estimation method of lithium battery State of Charge (SOC) is carried out, and for SOC estimation, a composite method of BP neural network with dynamic adaptive learning rate and Dynamic Adaptive Learning Rate BP Neural Network Modified Untraceable Kalman Filter (DABP-U) is proposed. The paper first employs the Unscented Kalman Filter (UKF) algorithm to make a preliminary estimation of the battery SOC, and then corrects and optimizes the UKF estimation results using a BP neural network with a dynamic adaptive learning rate, thereby improving the accuracy of the SOC estimation. The simulation and validation results show that the DABP-U method has higher estimation accuracy, stability, and faster convergence speed compared with the traditional algorithm, and especially performs well in dealing with the uncertainty of battery model parameters and the variation of the external environment.