Scholarlink

Wind power is a rapidly expanding source of renewable energy, yet it is significantly hampered by blade icing. Data-centric methodologies offer promising solutions for detecting blade icing. However, they typically necessitate the aggregation of substantial volumes of IoT data on a central server, potentially compromising sensitive business information. To overcome this limitation, this study introduces BLADE, an innovative blockchain-integrated federated learning (FL) framework tailored for blade icing detection in scenarios with imbalanced datasets. By leveraging blockchain technology, BLADE enhances conventional FL by mitigating reliance on a single centralized server and bolstering privacy protection. A novel validation mechanism, embedded within the Blockchain, fortifies defense against poisoning attacks. Furthermore, BLADE incorporates a unique imbalanced learning algorithm to effectively address the class imbalance inherent in sensor data. The efficacy of BLADE is rigorously assessed using data from 10 wind turbines across two distinct wind farms. Empirical findings substantiate the effectiveness, superiority, and practical applicability of the proposed BLADE framework.

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

Blockchain-Empowered Clustered Federated Convolutional Neural Network for Blade Icing Detection

  • 章节

  • 发表日期：

    2025年

查看全文
  • Xu Cheng,
  • Fan Shi,
  • Xiufeng Liu,
  • Shengyong Chen

摘要

Wind power is a rapidly expanding source of renewable energy, yet it is significantly hampered by blade icing. Data-centric methodologies offer promising solutions for detecting blade icing. However, they typically necessitate the aggregation of substantial volumes of IoT data on a central server, potentially compromising sensitive business information. To overcome this limitation, this study introduces BLADE, an innovative blockchain-integrated federated learning (FL) framework tailored for blade icing detection in scenarios with imbalanced datasets. By leveraging blockchain technology, BLADE enhances conventional FL by mitigating reliance on a single centralized server and bolstering privacy protection. A novel validation mechanism, embedded within the Blockchain, fortifies defense against poisoning attacks. Furthermore, BLADE incorporates a unique imbalanced learning algorithm to effectively address the class imbalance inherent in sensor data. The efficacy of BLADE is rigorously assessed using data from 10 wind turbines across two distinct wind farms. Empirical findings substantiate the effectiveness, superiority, and practical applicability of the proposed BLADE framework.