Scholarlink

<p>To address the problem of low efficiency in estimating the state of health (SOH) of lithium-ion batteries, a method based on the maximal information coefficient (MIC) algorithm and the back propagation (BP) neural network optimized by the golden jack optimization (GJO) algorithm is proposed in this study. Firstly, six aging features of SOH were extracted from the University of Maryland’s lithium-ion battery aging test data, and three high-quality aging features were selected using the MIC algorithm; then, the GJO algorithm is selected to optimize the initial weights and thresholds of the BP neural network to eliminate the problem of overfitting in the BP neural network; finally, GJO-BP was compared with BP neural networks optimized by genetic algorithm (GA) and simulated annealing (SA) algorithm. The results showed that after optimization using the MIC algorithm, the average error (MAE) of the model decreased by 31.37% compared to before optimization for aging characteristics; the reduction in MAE for GJO-BP compared to BP is 18.57% and 22.85% higher than that for GA-BP and SA-BP, respectively, while the convergence speed of GJO-BP is 50% faster than that of SA-BP. High-efficiency lithium battery SOH estimation can be achieved.</p>

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

State of health estimation of lithium-ion batteries based on maximal information coefficient feature optimization and GJO-BP neural network

  • 期刊论文

  • 发表日期：

    2025年2月13日

查看全文
  • Kou Farong,
  • Zhou Dongming,
  • Yang Tianxiang,
  • Luo Xi

摘要

To address the problem of low efficiency in estimating the state of health (SOH) of lithium-ion batteries, a method based on the maximal information coefficient (MIC) algorithm and the back propagation (BP) neural network optimized by the golden jack optimization (GJO) algorithm is proposed in this study. Firstly, six aging features of SOH were extracted from the University of Maryland’s lithium-ion battery aging test data, and three high-quality aging features were selected using the MIC algorithm; then, the GJO algorithm is selected to optimize the initial weights and thresholds of the BP neural network to eliminate the problem of overfitting in the BP neural network; finally, GJO-BP was compared with BP neural networks optimized by genetic algorithm (GA) and simulated annealing (SA) algorithm. The results showed that after optimization using the MIC algorithm, the average error (MAE) of the model decreased by 31.37% compared to before optimization for aging characteristics; the reduction in MAE for GJO-BP compared to BP is 18.57% and 22.85% higher than that for GA-BP and SA-BP, respectively, while the convergence speed of GJO-BP is 50% faster than that of SA-BP. High-efficiency lithium battery SOH estimation can be achieved.