多变量时序标记Transformer及其在电潜泵故障诊断中的应用  

Multivariate time series tokenized Transformer and its application in fault diagnosis of electric submersible pump

作　　者：李康[1,3] 李爽 高小永 李强[1] 张来斌[2,3] LI Kang;LI Shuang;GAO Xiao-yong;LI Qiang;ZHANG Lai-bin(College of Artificial Intelligence,China University of Petroleum(Beijing),Beijing 102249,China;College of Safety and Ocean Engineering,China University of Petroleum(Beijing),Beijing 102249,China;Key Laboratory of Oil and Gas Safety and Emergency Technology,Ministry of Emergency Management,China University of Petroleum(Beijing),Beijing 102249,China)

机构地区：[1]中国石油大学(北京)人工智能学院,北京102249 [2]中国石油大学(北京)安全与海洋工程学院,北京102249 [3]中国石油大学(北京)应急管理部油气生产安全与应急技术重点实验室,北京102249

出　　处：《控制与决策》2025年第4期1145-1153,共9页Control and Decision

基　　金：国家自然科学基金重点项目(52234007);国家自然科学基金面上项目(22178383);北京市自然科学基金面上项目(2232021);中国博士后科学基金项目(GZB20230861);中国石油大学(北京)科研基金项目(2462023YJRC036)。

摘　　要：电潜泵故障诊断对于确保安全可靠采油至关重要,但是,电潜泵数据呈现出的多变量、非线性和动态变化等复杂特性为该任务带来了严峻挑战.近年来,深度学习在复杂数据特征提取方面表现出的强大能力催生了一系列基于神经网络的电潜泵故障诊断方法.然而,多数方法忽略了电潜泵数据的动态特性以及长时依赖特征提取困难的问题.针对上述问题,提出一种多变量时序标记Transformer神经网络来实现电潜泵故障诊断.该模型设计新的多变量时间序列标记策略,继承引入多头注意力机制和残差连接的传统Transformer神经网络编码器在长时依赖特征提取方面的优势,用前向神经网络替代传统Transformer神经网络解码器来简化模型复杂度.通过对油田现场故障数据分析,验证所提出方法的有效性.实验结果表明,所提出方法实现了10类电潜泵故障的精确诊断,相比于流行的深度学习方法诊断性能更优.Fault diagnosis of electric submersible pumps is crucial to ensure safe and reliable oil recovery,but the complex characteristics of electric submersible pump data,such as multivariate,nonlinear,and dynamic changes,pose a great challenge to this task.In recent years,the powerful capabilities of deep learning in feature extraction of complex data have spawned a series of neural network-based fault diagnosis methods for electric submersible pumps.However,most of the methods ignore the dynamic characteristics of the electric submersible pump data as well as the difficulty of long-term dependency feature extraction.To address the above issues,a multivariate time-series tokenized Transformer neural network is proposed for fault diagnosis of electric submersible pumps.This model designs a new multivariate time series tokenization strategy and inherits the advantages of the traditional Transformer neural network’s encoder which introduces multi-head attention mechanisms and residual connections in long-term dependency feature extraction.In addition,this model replaces the traditional Transformer neural network’s decoder with a forward neural network to simplify the model’s complexity.The effectiveness of the method is verified by analyzing the field’s faulty data.The experimental results show that the proposed method can accurately diagnose ten types of electric submersible pump faults,and it outperforms the popular deep learning methods.

关 键 词：电潜泵 Transformer神经网络 深度学习 特征提取 故障诊断 多变量时序标记 

分 类 号：TP206.3[自动化与计算机技术—检测技术与自动化装置] TP18[自动化与计算机技术—控制科学与工程]