Scholarlink

<p>Energy storage batteries, as the core of energy storage technology, directly affect the overall efficiency and safe operation of new power systems through their performance and stability. In order to enhance the safety and reliability of energy storage batteries, this paper proposes a data-driven early fault warning method for energy storage batteries. Firstly, the self-attention mechanism (SAM) is employed to capture important information from the input sequence and assign different weights to it. Secondly, the DeepAR model is utilized to learn the trend of battery voltage changes, comprehensively capturing the state variations of the battery. Lastly, the local outlier factor (LOF) algorithm is adopted to build a battery fault early warning model, and grid search is employed to optimize the model’s efficiency and accuracy. Verified through actual data from an energy storage power station in Shanghai, the results indicate that the proposed model has an error within 0.16% when predicting voltage. Additionally, it achieves a 24-h early fault warning with an accuracy rate of 99.6%, providing reliable support for the safe and stable operation of energy storage systems.</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

Research on early fault warning for energy storage batteries based on data-driven approaches

  • 期刊论文

  • 发表日期：

    2025年7月24日

查看全文
  • Cheng Guixue,
  • Liang Ziyi,
  • Zhang Chao

摘要

Energy storage batteries, as the core of energy storage technology, directly affect the overall efficiency and safe operation of new power systems through their performance and stability. In order to enhance the safety and reliability of energy storage batteries, this paper proposes a data-driven early fault warning method for energy storage batteries. Firstly, the self-attention mechanism (SAM) is employed to capture important information from the input sequence and assign different weights to it. Secondly, the DeepAR model is utilized to learn the trend of battery voltage changes, comprehensively capturing the state variations of the battery. Lastly, the local outlier factor (LOF) algorithm is adopted to build a battery fault early warning model, and grid search is employed to optimize the model’s efficiency and accuracy. Verified through actual data from an energy storage power station in Shanghai, the results indicate that the proposed model has an error within 0.16% when predicting voltage. Additionally, it achieves a 24-h early fault warning with an accuracy rate of 99.6%, providing reliable support for the safe and stable operation of energy storage systems.