摘要
地区电网生产管理系统拥有大量的输变电设备缺陷记录文本,但是缺少系统化的利用和整理,可采用自然语言处理技术挖掘并分析缺陷记录文本。考虑电力系统领域专业分词方法,对缺陷记录文本中的句子进行文本预处理。利用word2vec模型将词语映射到高维特征空间,形成分布式词向量。提出基于注意力机制的双向长短期记忆神经网络(BiLSTM based on attention mechanism,BiLSTM-Attention)对电力设备缺陷文本进行分类,并以准确率、混淆矩阵以及微平均F1值(F1Micro)作为评价指标与传统机器学习模型、深度学习模型进行对比。实例分析表明BiLSTM-Attention模型具有更好的分类效果,能够实现电力设备缺陷文本的自动精准高效分类。最后构建电力设备缺陷处理优先指数,以此提升输变电设备的智能化管理水平,促进自然语言处理技术在电网中的应用。
There are a large number of defect records of power transmission and transformation equipment in regional power grid production management systems.The records lack systematic utilization and organization.Natural language processing can be used to mine and analyze defect record texts.This paper first considered the professional word segmentation method in the field of the power system and preprocesses the sentences in the defect record text.Then the word2 vec model is used to map the words to a high-dimensional feature space to form a distributed word vector.Finally,the bidirectional long short-term memory neural network based on attention mechanism(BiLSTM-Attention)was proposed to classify power equipment defect texts.The accuracy rate,confusion matrix,and F1 Micro value were used as evaluation indicators to compare with traditional machine learning methods and deep learning methods.The example analysis shows that the bidirectional long short-term memory neural network based on the attention mechanism has a better classification result,which can realize automatic,accurate and efficient classification of power equipment defect texts.In the last part,a priority index for processing power equipment defect was constructed to improve the management level of power transmission and transformation equipment and promote the application of natural language processing technology in the power grid.
作者
冯斌
张又文
唐昕
郭创新
王坚俊
杨强
王慧芳
FENG Bin;ZHANG Youwen;TANG Xin;GUO Chuangxin;WANG Jianjun;YANG Qiang;WANG Huifang(College of Electrical Engineering,Zhejiang University,Hangzhou 310027,Zhejiang Province,China;State Grid Zhejiang Maintenance Branch Company,Hangzhou 310027,Zhejiang Province,China)
出处
《中国电机工程学报》
EI
CSCD
北大核心
2020年第S01期1-10,共10页
Proceedings of the CSEE
基金
国家自然科学基金重点项目(51537010)
平湖市通用电气安装有限公司科技项目(2019-LHKJ-014)
关键词
缺陷记录
电力文本挖掘
注意力机制
双向长短期记忆网络
defect record
electric text mining
attention mechanism
bidirectional long short-term memory network(BiLSTM)
作者简介
冯斌(1997),男,博士研究生,研究方向为人工智能在电力系统中的应用,11910028@zju.edu.cn;张又文(1996),女,硕士研究生,研究方向为大数据与人工智能在电力系统中的应用,21810164@zju.edu.cn;唐昕(1980),男,博士,教授级高级工程师,研究方向为电网运维检修管理、综合能源管理,tang_xin@zju.edu.cn;郭创新(1969),男,教授,博士生导师,研究方向为电网智能调度控制与风险调度控制理论与应用、综合能源系统技术等,guochuangxin@zju.edu.cn;王坚俊(1973),男,高级工程师,研究方向为电力系统自动化、电网运行检修管理,hzdlwij@163.com;杨强(1979),男,教授,博士生导师,研究方向为综合能源系统规划与调控、电网自愈控制与网络安全,qyang@zju.edu.cn;王慧芳(1974),女,博士,副教授,研究方向为电力大数据挖掘,状态检修,电力系统继电保护及控制,huifangwang@zju.edu.cn。
