Scholarlink

<p> With the expansion of Energy Storage Power Stations (ESPS), the state assessment of Lithium-ion Batteries (LIBs) is crucial for system safety and efficiency. This study proposes a fusion algorithm combining adaptive extended Kalman filtering and particle swarm optimization to address traditional methods’ limitations in adapting to battery dynamic characteristics and reducing estimation errors. This algorithm dynamically adjusts the noise covariance matrix through an adaptive noise update mechanism, enhances the global search capability of particle swarm optimization, and makes the estimation results more accurate and reliable. Experiments showed the method’s loss values decreased to 0.1 and 0.06 across two datasets, with mean absolute errors in SOC estimation of only 0.98% and 0.62%. The identification error rapidly decreased with iterations, remaining between 0.2% and 0.3%. In practical applications, the method maintained battery SOC at 80%-90% under high-frequency low-power pulse conditions and long-term high-power continuous conditions with 4A current and approximately 1-second transient response. The designed state evaluation model effectively alleviates energy storage system pressure, reduces energy loss, and extends battery life, providing a new direction for LIBs state evaluation in ESPS and contributing to improved operational efficiency and safety.</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 evaluation of lithium-ion batteries in energy storage stations based on adaptive noise updating AEKF algorithm

  • 期刊论文

  • 发表日期：

    2026年1月10日

查看全文
  • Mingwan Zhuang,
  • Jianzhong Tang,
  • Junwei Ma,
  • Guanhui Yin,
  • Weirong Yang

摘要

With the expansion of Energy Storage Power Stations (ESPS), the state assessment of Lithium-ion Batteries (LIBs) is crucial for system safety and efficiency. This study proposes a fusion algorithm combining adaptive extended Kalman filtering and particle swarm optimization to address traditional methods’ limitations in adapting to battery dynamic characteristics and reducing estimation errors. This algorithm dynamically adjusts the noise covariance matrix through an adaptive noise update mechanism, enhances the global search capability of particle swarm optimization, and makes the estimation results more accurate and reliable. Experiments showed the method’s loss values decreased to 0.1 and 0.06 across two datasets, with mean absolute errors in SOC estimation of only 0.98% and 0.62%. The identification error rapidly decreased with iterations, remaining between 0.2% and 0.3%. In practical applications, the method maintained battery SOC at 80%-90% under high-frequency low-power pulse conditions and long-term high-power continuous conditions with 4A current and approximately 1-second transient response. The designed state evaluation model effectively alleviates energy storage system pressure, reduces energy loss, and extends battery life, providing a new direction for LIBs state evaluation in ESPS and contributing to improved operational efficiency and safety.