Spatio-temporal Graph Attention Network-based Fault Warning of Coal Mill

NIU Yuguang; XU Jian; DU Ming; YAO Jun

doi:10.19805/j.cnki.jcspe.2025.240423

English Version Reader LoginJournal Login

您当前的位置：

首页 >

文章列表页 >

Spatio-temporal Graph Attention Network-based Fault Warning of Coal Mill

更新时间：2025-11-28

- Spatio-temporal Graph Attention Network-based Fault Warning of Coal Mill
- Chinese Journal of Power Engineering Vol. 45, Issue 9, Pages: 1519-1526(2025)
- 作者机构：
  
  1. 华北电力大学新能源电力系统全国重点实验室,北京,102206
  2. 华北电力大学控制与计算机工程学院,北京,102206
  3. 北京怀柔实验室,北京,101400
- 作者简介：
- 基金信息：
- DOI：10.19805/j.cnki.jcspe.2025.240423
  CLC：
- Published Online：16 September 2025，
  
  Published：2025
- 稿件说明：
移动端阅览
牛玉广,徐健,杜鸣,姚珺. 基于时空图注意网络的磨煤机故障预警动力工程学报, 2025, 45(9): 1519-1526 https://doi.

org/10.19805/j.cnki.jcspe.2025.240423
牛玉广,徐健,杜鸣,姚珺. 基于时空图注意网络的磨煤机故障预警动力工程学报, 2025, 45(9): 1519-1526 https://doi. DOI： 10.19805/j.cnki.jcspe.2025.240423.

org/10.19805/j.cnki.jcspe.2025.240423 DOI：

摘要

为提升磨煤机故障预警结果的准确性与可信度

提出一种基于时空图注意网络的故障预警方法。通过最大信息系数与Top-K最近邻方法自适应求取邻接矩阵

将原始列表数据重构为时序图数据。随后

依次使用图注意力网络和双向门控循环单元

分别提取图数据的空间特征和时间特征

并对下一时刻数据进行预测。在离线阶段

经指数加权移动平均法计算总体预警阈值与各分量阈值。在在线阶段

当预测残差总量越限时发出预警信号

同时绘制各分量的越限分数热力图。结果表明:以某热电机组中速磨煤机运行数据为例

所提方法能够准确预警设备潜在异常

并对预警原因进行有效解释

优于对比方法。

Abstract

To improve the accuracy and reliability of fault warning for coal mills

a spatio-temporal graph attention network-based fault warning method was proposed. The maximal information coefficient and Top-K nearest neighbours were used to obtain the adjacency matrix adaptively. The original list data were reconstructed into sequence graph data. The spatial and temporal features of the graph data were extracted using the graph attention network and bidirectional gate recurrent unit in turn

and the next time data were predicted. In the offline stage

the overall warning threshold and each component threshold were calculated by the exponential weighted moving average method. In the online stage

a warning signal was issued when the total amount of prediction residuals exceeded the limit

meanwhile the heat map of the exceedance score of each component was plotted. Results show that taking the operation data of the medium-speed coal mill in a cogeneration unit as an example

the proposed method can accurately warn the potential anomalies of the equipment and effectively explain the reasons for the warning

which is better than relevant comparison methods.

关键词

Keywords

references

牛玉广, 李晓彬, 张佳辉. 基于多元状态估计与自适应闽值的电站辅机故障预警[J]. 动力工程学报, 2019, 39(9): 717-724. NIU Yuguang, LI Xiaobin, ZHANG Jiahui. Early fault warning for auxiliary equipment of power plants based on multivariate state estimation and adaptive threshold[J]. Journal of Chinese Society of Power Engineering, 2019, 39(9): 717-724.

杨婷婷, 李浩千, 陈晓峰, 等. 基于WPT-Transformer的磨煤机故障预警研究[J]. 热力发电, 2023, 52(12): 180-189. YANG Tingting, LI Haoqian, CHEN Xiaofeng, et al. Research on fault early warning of coal mill based on WPT and Transformer[J]. Thermal Power Generation, 2023, 52(12): 180-189.

HUANG Congzhi, QU Shuangyan, KE Zhiwu, et al. Dual fault warning method for coal mill based on Autoformer WaveBound[J]. Reliability Engineering & System Safety, 2024, 245: 110030.

姚天杨, 茅大钧, 韩万里. 基于DPC-MND和多元状态估计的磨煤机故障预警研究[J]. 控制工程, 2024, 31(5): 928-937. YAO Tianyang, MAO Dajun, HAN Wanli. Research on fault warning of coal mills based on DPC-MND and multivariate state estimation technique[J]. Control Engineering of China, 2024, 31(5): 928-937.

李晓彬, 牛玉广, 葛维春, 等. 基于改进堆叠自编码网络的电站辅机故障预警[J]. 仪器仪表学报, 2019, 40(6): 39-47. LI Xiaobin, NIU Yuguang, GE Weichun, et al. Early fault warning of power plant auxiliary engine based on improved stacked autoencoder network[J]. Chinese Journal of Scientific Instrument, 2019, 40(6): 39-47.

赵征, 丁建平. 基于深度双向门控循环神经网络的制粉系统故障预警[J]. 动力工程学报, 2023, 43(5): 598-605. ZHAO Zheng, DING Jianping. Fault warning of pulverizing system based on deep bidirectional gated recurrent neural network[J]. Journal of Chinese Society of Power Engineering, 2023, 43(5): 598-605.

罗毅, 段明达. 基于IWOA-Transformer的磨煤机故障预警[J]. 动力工程学报, 2024, 44(6): 939-946. LUO Yi, DUAN Mingda. Fault warning of coal mill based on IWOA-Transformer[J]. Journal of Chinese Society of Power Engineering, 2024, 44(6): 939-946.

赵玲, 龚加兴, 黄大荣, 等. 基于Fisher Score与最大信息系数的齿轮箱故障特征选择方法[J]. 控制与决策, 2021, 36(9): 2234-2240. ZHAO Ling, GONG Jiaxing, HUANG Darong, et al. Fault feature selection method of gearbox based on Fisher Score and maximum information coefficient[J]. Control and Decision, 2021, 36(9): 2234-2240.

LIU Jiayang, WANG Xiaosun, XIE Fuqi, et al. Condition monitoring of wind turbines with the implementation of spatio-temporal graph neural network[J]. Engineering Applications of Artificial Intelligence, 2023, 121: 106000.

侯磊, 刘金环, 于旭, 等. 图神经网络研究综述[J]. 计算机科学, 2024, 51(6): 282-298. HOU Lei, LIU Jinhuan, YU Xu, et al. Review of graph neural networks[J]. Computer Science, 2024, 51(6): 282-298.

VELIKOVI P, CUCURULL G, CASANOVA A, et al. Graph attention networks[C]//International Conference on Learning Representations (ICLR). Canada:[s.n.],2018.

唐贵基, 周威, 王晓龙, 等. 基于变维GRU-BiLSTM神经网络模型的滚动轴承寿命预测[J]. 中国工程机械学报, 2022, 20(6): 498-503. TANG Guiji, ZHOU Wei, WANG Xiaolong, et al. Residual life prediction of rolling bearing based on variable dimension GRU and BiLSTM network model[J]. Chinese Journal of Construction Machinery, 2022, 20(6): 498-503.

刘长良, 徐健, 王梓齐. 基于时空自编码网络的风电齿轮箱状态监测[J]. 信息与控制, 2023, 52(1): 84-92. LIU Changliang, XU Jian, WANG Ziqi. Condition monitoring of wind turbine gearbox based on spatial-temporal autoencoder network[J]. Information and Control, 2023, 52(1): 84-92.

徐健, 刘长良, 王梓齐, 等. 基于缺失数据填补的风电齿轮箱状态监测研究[J]. 仪器仪表学报, 2022, 43(9): 88-97. XU Jian, LIU Changliang, WANG Ziqi, et al. Research on condition monitoring of wind turbine gearbox based on missing data imputation[J]. Chinese Journal of Scientific Instrument, 2022, 43(9): 88-97.

Views

下载量

CSCD

Alert me when the article has been cited

提交

Tools

Publicity Resources

Numerical Simulation Study on Air-Powder Deviation and Adjustable Orifice Balancing Performance of Direct-fired Pulverizing System

Day-ahead Irradiance Forecasting Framework Based on CEEMDAN-KDBSCAN-Informer

Mooring-line Tension Prediction Based on Re-decomposition and Deep Learning

Related Author

NIE Changda

XIA Yongjun

WU Jianqun

HAN Jinke

ZENG Jiawei

JIANG Xudong

GUO Xiuwen

WANG Xiaoli

Related Institution

School of Energy and Environmental Engineering, Hebei University of Technology

Jiangtou Guohua Xinfeng Power Generation Co., Ltd.

JXIC Energy Tech. Research Institute

School of Automation, Central South University

School of Electrical & Information Engineering, Changsha University of Science & Technology

AI问答

⁰