基于深度学习的电力系统异常数据自动捕获方法  

Automatic abnormal data acquisition method for power system based on deep learning

作　　者：叶宽 杨博 朱戎 谢欢 赵蕾 YE Kuan;YANG Bo;ZHU Rong;XIE Huan;ZHAO Lei(State Grid Beijing Electric Power Research Institute,Beijing 100075,China)

机构地区：[1]国网北京电科院,北京100075

出　　处：《电子设计工程》2022年第9期162-165,170,共5页Electronic Design Engineering

摘　　要：为在电网应用环境中实现对异常传输电子量的精确化处理,提出基于深度学习的电力系统异常数据自动捕获方法。联合Caffe深度学习框架,清洗各类型电力数据资源,通过异常检测标签编码的方式,实现基于深度学习的电力系统异常数据检测。在此基础上,设置多层次的自动化协议栈架构,借助异常数据拷贝计划,建立必要的数据捕获映射条件,实现基于深度学习的电力系统异常数据自动捕获方法的顺利应用。对比实验结果表明,与机器学习型捕获手段相比,深度学习捕获法在单位时间内所能处理的异常传输电子数量值更大,而所需的消耗等待时间却相对更短,符合精确化处理异常传输电子量的实际应用需求。In order to achieve accurate processing of abnormal transmission electronic quantity in the application environment of power grid,an automatic acquisition method of abnormal data in power system based on deep learning is proposed. Combined with Caffe’s deep learning framework,various types of power data resources are cleaned,and abnormal data detection of power system based on deep learning is realized by means of abnormal detection label coding. On this basis,a multi-level automation protocol stack architecture is set up,and with the help of abnormal data copy plan,necessary data capture and mapping conditions are established to realize the smooth application of the automatic acquisition method of abnormal data in power system based on deep learning. The comparative experimental results show that,compared with the machine learning acquisition method,the deep learning acquisition method can deal with a larger number of abnormal transmitted electrons in unit time,while the consumption waiting time is relatively shorter,which meets the practical application requirements of accurate processing of abnormal transmitted electrons.

关 键 词：深度学习 电力系统 异常数据 自动捕获 Caffe框架 数据清洗 

分 类 号：TN911.7[电子电信—通信与信息系统]