基于视觉字符增强的电力调度故障预案匹配  

Power Dispatching Fault Plan Matching Based on Visual Character Enhancement

作　　者：籍雯媗 崔建业 冯斌 谷炜 郑翔 郭创新[1] JI Wenxuan;CUI Jianye;FENG Bin;GU Wei;ZHENG Xiang;GUO Chuangxin(College of Electrical Engineering,Zhejiang University,Hangzhou 310027,Zhejiang Province,China;State Grid Zhejiang Electric Power Company Limited,Hangzhou 310007,Zhejiang Province,China)

机构地区：[1]浙江大学电气工程学院,浙江省杭州市310027 [2]国网浙江省电力有限公司,浙江省杭州市310007

出　　处：《中国电机工程学报》2022年第15期5439-5447,共9页Proceedings of the CSEE

基　　金：国网浙江省电力有限公司金华供电公司科技项目(5211JH2000S2)。

摘　　要：电力系统中,调度控制中心会制定一系列的预案作为调度人员处理故障的参考和操作指南,然而,预案中包含的大量文本形式的知识目前仍需由人工进行反复记忆和查询,效率低,速度慢。为了解决这个问题,加快故障处理进程,该文提出一种用于自然语言推理的视觉特征增强型长短期记忆模型(enhanced long short-term memory model based on visual character-enhanced word embeddings for natural language inference,VCWE-ESIM)来实现故障现象与故障预案的自动匹配。该文从模型收敛速度、匹配准确性以及运算速度3方面将VCWE-ESIM模型与已有的文本匹配模型进行对比。算例分析表明,提出的VCWE-ESIM模型综合性能最优,能够准确快速地实现故障现象与故障预案的自动匹配,减少故障处置的时间,加快故障恢复的速度,防止由于故障处置不及时带来的严重后果,提高电网调度的智能化水平。In the power system,the dispatch control center will formulate a series of plans as a reference and operation guide for dispatchers to deal with faults.However,the large amount of knowledge in the form of text in the plans still needs to be memorized and queried by hand,which is inefficient and slow.In order to solve this problem and speed up the process of fault handling,an enhanced long short-term memory model based on visual character-enhanced word embedding for natural language inference(VCWE-ESIM)was proposed to realize automatic matching of fault phenomena and fault plans.The VCWE-ESIM model was compared with the existing text matching model from the aspects of convergence speed,matching accuracy,and operation speed.The case analysis shows that the comprehensive performance of the VCWE-ESIM model proposed in this paper is the best,which can realize the automatic matching between fault phenomena and fault plans accurately and quickly.Thereby,it can reduce the time for fault handling,speed up recovery pace as well as prevent serious consequences caused by untimely fault handling,which can further improve the intelligent level of power grid dispatching.

关 键 词：故障预案 文本匹配 视觉特征 视觉特征增强型长短期记忆模型 

分 类 号：TM73[电气工程—电力系统及自动化]