异常检测
信息物理系统
传感器融合
异常(物理)
计算机科学
数据挖掘
电力系统
功率(物理)
人工智能
物理
凝聚态物理
操作系统
量子力学
作者
Tianlei Zang,Shijun Wang,Chuangzhi Li,Yunfei Liu,Yujian Xiao,Zian Wang,Xueying Yu
标识
DOI:10.1016/j.ijepes.2025.110813
摘要
The localized faults are easier to propagate across domains and escalate into cascading failures in cyber physical power system (CPPS) with the deep integration of cyber and physical components. As a result, the risks of CPPS have increased significantly. It is a challenge to fully quantify the complex characteristics of CPPS. A cyber-physical bilateral data-driven composite model is proposed in this paper to achieve efficient and accurate anomaly detection of CPPS. The novel model can depict data decomposition and feature extraction from both cyber and physical domains. First, a sample convolution and interaction network is built to effectively capture temporal dependencies and sudden anomaly features in physical-side data. The time-sensitive patterns and unique deviations are focused on ensuring accurate detection of physical-side anomalies. Second, a transformer-based detection model is established to extract dynamic network attributes and state transition patterns in cyber-side data. By accurately tracking evolving network behaviors and subtle state transitions, robust detection of anomalies in the cyber domain is ensured. Last, the extracted features from both cyber and physical domains are integrated into a unified representation to achieve cross-domain synergy to precisely map CPPS anomalies. Case studies on the IEEE 33-bus system validate the effectiveness and superior performance of proposed method in identifying diverse anomaly states and enhancing CPPS operational safety and stability.
科研通智能强力驱动
Strongly Powered by AbleSci AI