《计算机应用》唯一官方网站 ›› 2026, Vol. 46 ›› Issue (10): 3391-3399.DOI: 10.11772/j.issn.1001-9081.2025080937

• 前沿与综合应用 • 上一篇    

基于多头注意力和新门控单元的发电厂给水泵故障预警

王井阳1, 王浩1, 张宾2(), 时薇3(), 黄敏1, 张杨1   

  1. 1.河北科技大学 信息科学与工程学院,石家庄 050018
    2.河北工程技术学院 网络空间安全学院 石家庄 050091
    3.河北传媒学院 人工智能学院,石家庄 051430
  • 收稿日期:2025-08-15 修回日期:2025-09-01 接受日期:2025-09-01 发布日期:2026-10-21 出版日期:2026-10-10
  • 通讯作者: 张宾,时薇
  • 作者简介:王井阳(1971—),男,河北唐山人,教授,硕士,CCF高级会员,主要研究方向:机器学习、深度学习、人工智能
    王浩(1997—),男,安徽淮北人,硕士研究生,主要研究方向:深度学习、电力设备故障预警
    黄敏(1979—),男,河北邯郸人,教授,博士,CCF会员,主要研究方向:人工智能、深度学习、机器学习
    张杨(1980—),男,河北秦皇岛人,教授,博士,主要研究方向:软件重构、并发软件。
  • 基金资助:
    河北省自然科学基金资助项目(F2023208001)

Fault warning of feedwater pumps in power plant based on multi-head attention and new gated unit

Jingyang WANG1, Hao WANG1, Bin ZHANG2(), Wei SHI3(), Min HUANG1, Yang ZHANG1   

  1. 1.School of Information Science and Engineering,Hebei University of Science and Technology,Shijiazhuang Hebei 050018,China
    2.School of Cyberspace Security,Hebei University of Engineering Science,Shijiazhuang Hebei 050091,China
    3.School of Artificial Intelligence,Hebei Institute of Communications,Shijiazhuang Hebei 051430,China
  • Received:2025-08-15 Revised:2025-09-01 Accepted:2025-09-01 Online:2026-10-21 Published:2026-10-10
  • Contact: Bin ZHANG, Wei SHI
  • About author:WANG Jingyang, born in 1971, M. S., professor. His research interests include machine learning, deep learning, artificial intelligence.
    WANG Hao, born in 1997, M. S. candidate. His research interests include deep learning, early warning of power equipment failures.
    HUANG Min, born in 1979, Ph. D., professor. His research interests include artificial intelligence, deep learning, machine learning.
    ZHANG Yang, born in 1980, Ph. D., professor. His research interests include software refactoring, concurrent software.
  • Supported by:
    Hebei Provincial Natural Science Foundation(F2023208001)

摘要:

为实现发电厂给水泵故障的早期预警,提出一种基于多头注意力(MHA)机制和新门控单元(NGU)的给水泵故障预警模型——MHA-NGU。MHA用于捕捉时间序列中的长距离依赖关系,通过多视角特征交互强化关键维度;本文提出的NGU是一种改进循环神经网络(RNN),将它的输出门和重置门设置为两级,避免单级门因过度依赖局部信息导致的全局依赖捕捉能力下降,并在内部嵌入T转换模块,用于进一步缓解模型梯度消失问题。以浙江省某1 030 MW火电机组的给水泵为例进行验证,实验结果表明:所提模型比门控循环单元(GRU)和次优模型MHA-GRU的决定系数(R2)分别提高4.38%和1.90%,且比GRU提前5.39 h触发预警,有效实现了给水泵的早期故障预警。

关键词: 门控循环单元, 故障预警, 多头注意力机制, T转换模块, 循环神经网络, 时间序列预测

Abstract:

To achieve early fault warning for feedwater pumps in power plants, a feedwater pump fault warning model, MHA-NGU, based on Multi-Head Attention (MHA) mechanism and New Gated Unit (NGU) was proposed. MHA mechanism was employed to capture long-range dependencies in the time series data, thereby enhancing key dimensions through multi-perspective feature interaction. NGU is an improved Recurrent Neural Network (RNN) proposed in this paper, in which the output gate and reset gate were designed with two levels, thereby avoiding the degradation in global dependency capture capability caused by single?level gates’ over?reliance on local information. Additionally, a T-transformation module was embedded within NGU to further alleviate the gradient vanishing problem of the model. Experimental validation was conducted using data from a 1 030 MW thermal power unit in Zhejiang Province as an example. The results demonstrate that the proposed model increases the Coefficient of Determination (R2) by 4.38% and 1.90% compared with GRU and the second-best model, MHA-GRU, respectively. Furthermore, it triggers a warning 5.39 hours earlier than the GRU model, realizing an early fault warning for feedwater pumps effectively.

Key words: Gated Recurrent Unit (GRU), fault warning, Multi-Head Attention (MHA) mechanism, T-transformation module, Recurrent Neural Network (RNN), time series prediction

中图分类号: