基于双分支结构的电网外部入侵预警识别方法  

Early Warning Identification Method for External Instrution of Power Grid Based on Bilateral-branch Structure

作　　者：王南 徐世超 欧阳超 高峰[1,3] 魏昊焜[1] 马磊 杨彪 WANG Nan;XU Shichao;OUYANG Chao;GAO Feng;WEI Haokun;MA Lei;YANG Biao(State Grid Shaanxi Electric Power Research Institute,Xi’an 710100,China;School of Computer Science and Technology,Harbin Institute of Technology(Shenzhen),Shenzhen 518055,China;Xi’an Jiaotong University,Xi’an 710048,China;Jiangsu FUDI Electric Technology Co.Ltd.,Nantong 226253,China;School of Architecture,Harbin Institute of Technology(Shenzhen),Shenzhen 518055,China)

机构地区：[1]国网陕西电力科学研究院,陕西西安710100 [2]哈尔滨工业大学(深圳)计算机科学与技术学院,广东深圳518055 [3]西安交通大学,陕西西安710048 [4]江苏复迪电气科技有限公司,江苏南通226253 [5]哈尔滨工业大学(深圳)建筑学院,广东深圳518055

出　　处：《智慧电力》2022年第6期35-41,78,共8页Smart Power

基　　金：国家自然科学基金资助项目(61972112)。

摘　　要：针对电力网络场景下外部入侵目标的预警识别问题,提出了一种基于双分支结构的预警识别方法。首先,利用加权重采样分支提高对尾部类的识别准确率。其次,提出动态特征集成模块解决目标尺度大小不一的问题,并基于Sigmoid函数的分支融合策略,有效地融合尾部类特征和随机采样分支提取的通用特征。最后,使用数据集验证了所提方法分类的有效性。Aiming at the problem of early warning identification for intrusive objects in power grid scenarios,an early warning identification method based on bilateral-branch structure is proposed.Firstly,weighted re-sampling branch is used to enhance the recognition accuracy on tail class.Then,dynamic feature integration module is proposed to solve the problem of different target scales.Sigmoid function is used to effectively combine the features from tail class and the general features extracted by random sampling branch.Finally,the effectiveness of the proposed method is validated on our dataset.

关 键 词：电力网络 外部入侵目标 预警识别 双分支结构 

分 类 号：TM769[电气工程—电力系统及自动化]