基于DA-Xception算法的电容式电压互感器状态异常识别  

Abnormal State Identification of Capacitor Voltage Transformer Based on DA-Xception Algorithm

作　　者：姜瀚书 李雨田 吴广昊 刘著 郑永明 JIANG Hanshu;LI Yutian;WU Guanghao;LIU Zhu;ZHENG Yongming(State Grid Jilin Electric Power Co.,Ltd.,Marketing Service Center,Changchun 130062,China)

机构地区：[1]国网吉林省电力有限公司营销服务中心,吉林长春130062

出　　处：《电工技术》2024年第9期113-117,共5页Electric Engineering

摘　　要：为准确在线识别出电容式电压互感器是否出现异常状态,提出了一种融合DANet双重注意力机制的DA-Xception算法,对电容式电压互感器是否出现异常状态进行识别诊断。该算法将DANet双重注意力融合到Xception算法的输出流阶段,对提取到的特征通道信息和空间信息的相互依赖性进行学习,得到更符合目标上下文关系的有效特征。实验结果表明,DA-Xception算法对电容式电压互感器异常状态识别的准确率达到了97.7%,能对出现异常的电容式电压互感器进行高效、准确的判断识别。In order to accurately online identify whether a capacitor voltage transformer is undergoing abnormal states,a DA-Xception algorithm integrating DANet dual attention mechanism is proposed.This algorithm integrates the dual attention of DANet into the output stream stage of the Xception algorithm,learning the interdependence between extracted feature channel information and spatial information,and obtaining effective feature results that are more consistent with the contextual relationship of the target feature.The experimental results show that the DA-Xception algorithm can efficiently and accurately identify abnormal states of capacitor voltage transformers with an accuracy of 97.7%.

关 键 词：电容式电压互感器 异常识别 Xception DANet 双重注意力 

分 类 号：TM451[电气工程—电器]