Data source authentication of synchrophasor measurement devices based on 1D-CNN and GRU
Data source authentication of synchrophasor measurement devices based on 1D-CNN and GRU
复制标题
DOI：
10.1016/j.epsr.2021.107207
复制
发表时间：
2021-07
影响因子：
3.9
通讯作者：
Shengyuan Liu;Shutang You;Chujie Zeng;H. Yin;Zhenzhi Lin;Yuqing Dong;W. Qiu;Wenxuan Yao;Yilu Liu
中科院分区：
文献类型：
--
作者：
Shengyuan Liu;Shutang You;Chujie Zeng;H. Yin;Zhenzhi Lin;Yuqing Dong;W. Qiu;Wenxuan Yao;Yilu Liu
同步相量测量设备（SMD）已被广泛部署，以支持电力系统的实时监测和控制。与此同时，近年来出现了数据欺骗。因此，研究数据认证算法对于有效地检测和防御数据欺骗具有重要意义。在这项工作中，利用一维卷积神经网络（1D-CNN）提取隐藏在频率，电压角度和幅度数据中的时间签名;然后门控递归单元（GRU）采用这些时间签名进行数据源认证。在算例分析中，首次在具有多个SMD的大规模电力系统中对不同算法的性能进行了测试，并对不同算法进行了比较，结果表明，该算法能够在较短的时间窗口内实现较高的数据源认证精度。
Synchrophasor measurement devices (SMDs) have been widely deployed to support real-time monitoring and control of power systems. In the meantime, data spoofing has emerged in recent years. Therefore, it is of great importance to study data authentication algorithms for detecting and defending the data spoofing effectively. In this work, a one-dimensional convolutional neural network (1D-CNN) is utilized to extract temporal signatures hidden in frequency, voltage angle and amplitude data; then the gated recurrent unit (GRU) employs these temporal signatures for data source authentication. In case studies, the performances of different algorithms are tested in large-scale power systems with numerous SMDs for the first time, and comparisons among different algorithms show that the proposed algorithm can achieve a higher accuracy of data source authentication with a shorter time window.