Scholarlink

<p>Accurate assessment of the state of health (SOH) is essential for the safe, economical, and reliable operation of all-vanadium redox flow batteries (VRFBs). However, systems for monitoring their SOH have rarely been studied. In this study, capacity retention rate is used to characterize the SOH, and a series of charging/discharging cycling tests is conducted. Based on a series of operational parameters and grey relation analysis (GRA), the best health feature subset to characterize the SOH is determined. Extreme learning machine (ELM) combined with swarm intelligence algorithms is used to predict SOH, based on which the effects of different swarm intelligence algorithms and different activation functions on estimations of SOH performance are investigated. Results show that all the models with poor prediction results have a sigmoid activation function, indicating that such functions are not capable of predicting SOH. Grey relational analysis based on the LASSO (LGRA) algorithm-ELM is more suitable for SOH prediction than the interaction measurement algorithm based on Spearman’s correlation coefficient (SCIM) algorithm-ELM model (here, LASSO is the least absolute shrinkage and selection operator). It can not only estimate SOH accurately but also predict recession characteristics. The most suitable swarm intelligence algorithm is particle swarm optimization (PSO), and the best activation function is linear, in which the maximum error is lower than 0.2%.</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

Model for predicting the state of health of redox flow batteries based on an extreme learning machine model and swarm intelligence algorithm

  • 期刊论文

  • 发表日期：

    2026年5月8日

查看全文
  • Lei Zhang,
  • Zebo Huang,
  • Guo’an Yang,
  • Fengming Chu

摘要

Accurate assessment of the state of health (SOH) is essential for the safe, economical, and reliable operation of all-vanadium redox flow batteries (VRFBs). However, systems for monitoring their SOH have rarely been studied. In this study, capacity retention rate is used to characterize the SOH, and a series of charging/discharging cycling tests is conducted. Based on a series of operational parameters and grey relation analysis (GRA), the best health feature subset to characterize the SOH is determined. Extreme learning machine (ELM) combined with swarm intelligence algorithms is used to predict SOH, based on which the effects of different swarm intelligence algorithms and different activation functions on estimations of SOH performance are investigated. Results show that all the models with poor prediction results have a sigmoid activation function, indicating that such functions are not capable of predicting SOH. Grey relational analysis based on the LASSO (LGRA) algorithm-ELM is more suitable for SOH prediction than the interaction measurement algorithm based on Spearman’s correlation coefficient (SCIM) algorithm-ELM model (here, LASSO is the least absolute shrinkage and selection operator). It can not only estimate SOH accurately but also predict recession characteristics. The most suitable swarm intelligence algorithm is particle swarm optimization (PSO), and the best activation function is linear, in which the maximum error is lower than 0.2%.