基于注意力机制的CNN-BiLSTM操作票自动校核方法  

Automatic verification method of operation ticket based on attention mechanism of CNN-BiLSTM

作　　者：周凯 焦龄霄 胡志坚[2] 严利雄 毕如玉 王勇杰 ZHOU Kai;JIAO Lingxiao;HU Zhijian;YAN Lixiong;BI Ruyu;WANG Yongjie(Extra High Voltage Company,State Grid Hubei Electric Power Co.,Ltd.,Wuhan 430050,China;School of Electrical Engineering and Automation,Wuhan University,Wuhan 430072,China)

机构地区：[1]国网湖北省电力有限公司超高压公司,湖北武汉430050 [2]武汉大学电气与自动化学院,湖北武汉430072

出　　处：《武汉大学学报（工学版）》2023年第9期1114-1123,共10页Engineering Journal of Wuhan University

基　　金：国家自然科学基金项目(编号:51977156);国网湖北电力公司科技项目(编号:52152020003K)。

摘　　要：针对常规电力系统操作票的安全校验依赖人工经验、主观性强、容易出错、可靠性不高且校验效率较低,提出了一种基于注意力机制的CNN(convolutional neural network)-BiLSTM(bidirectional long short-term memory network)操作票自动校核方法。该方法首先对操作票文本进行分词处理,并利用文本向量化模型将操作票文本转变为词向量矩阵;然后以词向量矩阵作为CNN的输入,提取操作票文本局部词语间的高维语义特征,以高维语义特征构成的序列作为BiLSTM网络的输入,进一步挖掘操作票文本的上下文联系;同时引入注意力机制给予BiLSTM网络隐藏层不同的权重以加强关键词语的影响,最终实现操作票文本的深度语义特征挖掘,通过全连接层将深度语义特征映射到校核标签空间,给出调度操作票文本的校核判定结果。以湖北某地区电网的操作票为样本进行实验,实验结果表明,该方法校验正确率较高,能够较为准确地判别操作票的正确性,有效提高操作票校验的工作效率。The conventional verification of power system operation ticket relies on manual experience,which is subjective and error-prone,low reliability and low verification efficiency.Aiming at the above-mentioned problems,an automatic verification method of operation ticket of CNN(convolutional neural network)-BiLSTM(bidirectional long short-term memory network)based on attention mechanism is proposed.This method first performs word segmentation processing on the operation ticket text,and then uses the text vectorization model to transform the operation ticket text into a word vector matrix.Then,the word vector matrix is taken as the input of the CNN to extract high-dimensional semantic features between local words in the operation ticket,and the series of high-dimensional semantic features are used as the input of the BiLSTM network to further explore the contextual relationship of the operation ticket text.At the same time,the attention mechanism is introduced to give different weights to the hidden layer of the BiLSTM network to strengthen the influence of keywords,and finally realize the deep semantic feature mining of the operation ticket text.Through the fully connected layer,the deep semantic features are mapped to the check label space,and the check judgment result of the dispatch operation ticket text is given.Experiments are carried out by taking the operation ticket of a certain power grid in Hubei Province as a sample.The experimental results show that the method has a higher verification rate,which can more accurately determine the correctness of the operation ticket and effectively improve the work efficiency of operation ticket verification.

关 键 词：调度操作票 卷积神经网络 双向长短期记忆网络 注意力机制 文本向量化 

分 类 号：TM631[电气工程—电力系统及自动化]