Scholarlink

<p>This study proposes a novel smart grid intrusion detection model, combining a quantum-enhanced beetle swarm optimization algorithm with extreme learning machine (QBOA-ELM), with the aim of improving detection accuracy, efficiency, and robustness. By integrating quantum-enhanced optimization strategies, QBOA-ELM demonstrates superior performance across multiple aspects. The experimental results demonstrate that QBOA-ELM achieves significantly better accuracy (97.5%), recall (96.8%), precision (97.2%), F1-Score (0.972), sensitivity (96.8%), and specificity (98.1%) in intrusion detection tasks when compared to traditional Beetle Swarm Optimization Extreme Learning Machine (BOA-ELM) and other classical algorithms such as Support Vector Machine (SVM) and Decision Tree (DT). On the standard smart grid dataset, QBOA-ELM outperforms BOA-ELM (94.3% accuracy, 92.5% recall) and surpasses SVM (91.7%) and DT (89.5%). With regard to the efficiency of training, QBOA-ELM demonstrates a substantial advantage, processing large-scale datasets in a training time of 120&#xa0;s, which is considerably less than the 180&#xa0;s required by BOA-ELM and the 300&#xa0;s required by SVM. Furthermore, after dimensionality reduction via principal component analysis (PCA), QBOA-ELM maintains high performance on high-dimensional datasets, achieving accuracy and recall rates of 97.2% and 96.5%, respectively, outperforming other models. Furthermore, the model demonstrates strong robustness in the presence of noise interference (10% and 20% random noise), with accuracy and recall rates of 96.8% and 95.9%, respectively, far exceeding traditional methods. Consequently, the efficacy of QBOA-ELM is twofold: firstly, it enhances intrusion detection performance, and secondly, it offers advantages in real-time and large-scale data processing. This renders it well-suited for the security monitoring of complex systems such as smart grids. Future work could integrate deep learning techniques to further optimise the model and extend its applications.</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

Quantum-enhanced beetle swarm optimized ELM for high-dimensional smart grid intrusion detection

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年7月26日

查看全文
  • Na Cheng,
  • Shuqing Wang,
  • Lihong Zhao,
  • Yan Hu

摘要

This study proposes a novel smart grid intrusion detection model, combining a quantum-enhanced beetle swarm optimization algorithm with extreme learning machine (QBOA-ELM), with the aim of improving detection accuracy, efficiency, and robustness. By integrating quantum-enhanced optimization strategies, QBOA-ELM demonstrates superior performance across multiple aspects. The experimental results demonstrate that QBOA-ELM achieves significantly better accuracy (97.5%), recall (96.8%), precision (97.2%), F1-Score (0.972), sensitivity (96.8%), and specificity (98.1%) in intrusion detection tasks when compared to traditional Beetle Swarm Optimization Extreme Learning Machine (BOA-ELM) and other classical algorithms such as Support Vector Machine (SVM) and Decision Tree (DT). On the standard smart grid dataset, QBOA-ELM outperforms BOA-ELM (94.3% accuracy, 92.5% recall) and surpasses SVM (91.7%) and DT (89.5%). With regard to the efficiency of training, QBOA-ELM demonstrates a substantial advantage, processing large-scale datasets in a training time of 120 s, which is considerably less than the 180 s required by BOA-ELM and the 300 s required by SVM. Furthermore, after dimensionality reduction via principal component analysis (PCA), QBOA-ELM maintains high performance on high-dimensional datasets, achieving accuracy and recall rates of 97.2% and 96.5%, respectively, outperforming other models. Furthermore, the model demonstrates strong robustness in the presence of noise interference (10% and 20% random noise), with accuracy and recall rates of 96.8% and 95.9%, respectively, far exceeding traditional methods. Consequently, the efficacy of QBOA-ELM is twofold: firstly, it enhances intrusion detection performance, and secondly, it offers advantages in real-time and large-scale data processing. This renders it well-suited for the security monitoring of complex systems such as smart grids. Future work could integrate deep learning techniques to further optimise the model and extend its applications.