Scholarlink

<p>Accurate estimation of lithium-ion battery (LIB) remaining useful life (RUL) based on degradation data is critical for advancing battery performance optimization, yet it remains challenging due to inherent nonlinear degradation trends and prominent capacity regeneration phenomena. To address this, this study proposes a novel hybrid framework integrating complementary ensemble empirical mode decomposition (CEEMD), adaptive unscented Kalman filter (AUKF), and sparrow search algorithm-optimized gated recurrent unit (SSA-GRU) for RUL estimation. The framework features three synergistic innovations: (1) CEEMD is employed to effectively mitigate capacity regeneration and noise interference, outperforming conventional decomposition methods in preserving degradation signal integrity; (2) a double exponential state space model is constructed to characterize LIB degradation dynamics, with AUKF enabling robust prediction of global degradation trends; (3) SSA-GRU is introduced to specifically correct AUKF prediction errors and capacity regeneration, realizing high-precision multi-stage capacity prediction. Validated on NASA and CALCE Public datasets, the framework exhibits superior performance across 30% and 50% training data scenarios compared to existing single/multi-step prediction algorithms. Experimental results demonstrate its capability to capture both global trends and local characteristics of capacity degradation, with mean absolute error and root mean square error both below 1.62%, relative accuracy exceeding 96.5%, and 99.5% of capacity errors confined within 0.02 Ah.</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

Multi-stage capacity trajectory prediction for lithium-ion battery RUL estimation: CEEMD-AUKF-SSA-GRU fusion model and validation

  • 期刊论文

  • 发表日期：

    2025年10月9日

查看全文
  • Kangping Gao,
  • Jinxuan Lin,
  • Chengyao Liu,
  • Jianjie Sun,
  • Xinxin Xu

摘要

Accurate estimation of lithium-ion battery (LIB) remaining useful life (RUL) based on degradation data is critical for advancing battery performance optimization, yet it remains challenging due to inherent nonlinear degradation trends and prominent capacity regeneration phenomena. To address this, this study proposes a novel hybrid framework integrating complementary ensemble empirical mode decomposition (CEEMD), adaptive unscented Kalman filter (AUKF), and sparrow search algorithm-optimized gated recurrent unit (SSA-GRU) for RUL estimation. The framework features three synergistic innovations: (1) CEEMD is employed to effectively mitigate capacity regeneration and noise interference, outperforming conventional decomposition methods in preserving degradation signal integrity; (2) a double exponential state space model is constructed to characterize LIB degradation dynamics, with AUKF enabling robust prediction of global degradation trends; (3) SSA-GRU is introduced to specifically correct AUKF prediction errors and capacity regeneration, realizing high-precision multi-stage capacity prediction. Validated on NASA and CALCE Public datasets, the framework exhibits superior performance across 30% and 50% training data scenarios compared to existing single/multi-step prediction algorithms. Experimental results demonstrate its capability to capture both global trends and local characteristics of capacity degradation, with mean absolute error and root mean square error both below 1.62%, relative accuracy exceeding 96.5%, and 99.5% of capacity errors confined within 0.02 Ah.