|本期目录/Table of Contents|

[1]甘 李,姚 智,李 闯,等.基于卷积神经网络的汽轮机抗燃油泄漏智能预警技术研究[J].工业仪表与自动化装置,2022,(04):8-13+98.[doi:10.19950/j.cnki.cn61-1121/th.2022.04.002]
　GAN Li,YAO Zhi,LI Chuang,et al.Research on intelligent early warning technology of steam turbine anti fuel leakage based on convolutional neural network[J].Industrial Instrumentation & Automation,2022,(04):8-13+98.[doi:10.19950/j.cnki.cn61-1121/th.2022.04.002]
点击复制

基于卷积神经网络的汽轮机抗燃油泄漏智能预警技术研究

《工业仪表与自动化装置》[ISSN:1000-0682/CN:61-1121/TH]

卷:
期数:
2022年04期
页码:
8-13+98
栏目:
出版日期:
2022-08-15

文章信息/Info

Title:
Research on intelligent early warning technology of steam turbine anti fuel leakage based on convolutional neural network
文章编号:
1000-0682（2022）04-0000-00
作者:
甘 李1; 姚 智2; 李 闯1; 郭云飞2; 蔺奕存2; 李 昭2; 谭祥帅2; 王 林2
1.京能十堰热电有限公司，湖北 十堰 442000；
2.西安热工研究院有限公司，陕西 西安 710054
Author(s):
GAN Li1; YAO Zhi2; LI Chuang1; GUO Yunfei2; LIN Yicun2; LI Zhao2; TAN Xiangshuai2; WANG Lin2
1.Jingneng Shiyan Thermal Power Co., Ltd., Hubei Shiyan ,442000, China；
2.Xi’an Thermal Power Research Institute Co., Ltd., Shaanxi Xi’an 710032, China
关键词:
汽轮机; EH油系统; 卷积神经网络; 数据驱动; 火力发电厂
Keywords:
turbine;  EH oil system;  convolutional neural network;  data driven;  thermal power plant
分类号:
中图分类号：TM621.2
DOI:
10.19950/j.cnki.cn61-1121/th.2022.04.002
文献标志码:
A
摘要:
汽轮机数字电液控制系统依靠抗燃油（EH油）对汽门进行调节，抗燃油一旦发生泄漏，将危及机组的运行安全。为提高EH油泄漏故障的监测可靠性，减少电厂非计划停机次数，基于大数据挖掘与数据驱动技术，构建了多个不同结构的卷积神经网络模型，用于预测监控EH油箱油位。利用机组实际运行积累的大量数据，建立模型训练与验证数据集，从预测准确率、交叉熵损失及运算耗时等方面对模型进行综合评价。评估结果表明，A结构的模型准确性最高，约为98.92%，交叉熵损失最低，约为0.043 1，而模型运算时长中等，综合性能最优。将A智能模型整合进电厂分散式控制系统中进行实际验证，其监测准确，预警及时，显著减少了监盘和巡检人员的工作量，提高了电力生产的自动化、智能化水平，相关应用经验可供同类型机组参考。
Abstract:
The digital electro-hydraulic control system of steam turbine relies on fire-resistant oil (EH oil) to regulate the steam valve. Once the fire-resistant oil leaks, it will endanger the operation safety of the unit. In order to improve the monitoring reliability of EH oil leakage fault and reduce the number of unplanned shutdown of power plant, several convolution neural network models with different structures are constructed based on big data mining and data-driven technology to predict and monitor the oil level of EH oil tank. Using a large amount of data accumulated in the actual operation of the unit, the model training and verification data set is established, and the model is comprehensively evaluated from the aspects of prediction accuracy, cross entropy loss and operation time. The evaluation results show that the model accuracy of structure a is the highest, about 98.92%, and the cross entropy loss is the lowest, about 0.0431. The operation time of the model is medium, and the comprehensive performance is the best. The intelligent model a is integrated into the distributed control system of the power plant for practical verification. Its monitoring is accurate and early warning is timely, which significantly reduces the workload of supervision and inspection personnel, and improves the automation and intelligence level of power production. The relevant application experience can be used as a reference for similar units.

参考文献/References:

[1]王杭州.300 MW汽轮机电液调节系统故障分析与对策[J].发电设备,2003(01):19-22.

[2]马勇,杜文斌,薛志恒.330 MW机组EH油压偏低原因分析[J].华电技术,2012,34(03):38-40+48+83-84.
[3]GUO Jing-xiao, XIE Dan-mei. A Comparative Study on Oil System for Supercritical Steam Turbine Made in Domestic with 600MW Capacity[J]. Advanced Materials Research, 2012, 1566(433-440):4390-4395.
[4]LI W , LI X , NIU Q , et al. Abnormal shutdown analysis and treatment of sub-critical 300 MW unit due to low EH oil pressure[J]. IOP Conference Series Earth and Environmental Science, 2019, 300:042012.
[5]LI Wanjun,NIU Q,HUANG T. Shutdown analysis and treatment of 150MW steam extraction heating unit due to low oil fuel level[J]. IOP Conference Series: Earth and Environmental Science, 2019, 300:042014.
[6]王大伟.伺服阀常见故障的分析与处理[J].电力建设,2001(09):46-48.
[7]张豪.基于CNN的恶意软件分类方法[J].计算机时代,2021(12):48-51.
[8]戴彦文,于艾清.基于健康特征参数的CNN-LSTM&GRU组合锂电池SOH估计[J/OL].储能科学与技术:1-9[2021-12-27].DOI:10.19799/j.cnki.2095-4239.2021.0623.
[9]张琳,易卿武,黄璐,等.基于Wavelet-CNN网络的人类活动识别技术[J/OL].无线电工程:1-9[2021-12-27].http://kns.cnki.net/kcms/detail/13.1097.TN.20211221.1245.008.html.
[10]LECUN Y,BOTTOU L. Gradient-based learning applied to document recognition[J]. Proceedings of the IEEE, 1998, 86(11) : 2278-2324.
[11]HE K , ZHANG X , REN S , et al. Spatial Pyramid Pooling in Deep Convolutional Networks for Visual Recognition[J]. CoRR, 2014, abs/1406.4729.
[12]高薪,胡月,杜威,等.腐蚀膨胀算法对灰度图像去噪的应用[J].北京印刷学院学报,2014,22(04):63-65.
[13]IOFFE S，SZEGEDY C. Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift[J]. JMLR.org,2015.
[14]JIA Y , SHELHAMER E , DONAHUE J , et al. Caffe: Convolutional Architecture for Fast Feature Embedding[J].ACM, 2014.
[15]郭智,宋萍,张义,等.基于深度卷积神经网络的遥感图像飞机目标检测方法[J].电子与信息学报,2018,40(11):2684-2690.
[16]温煌璐.基于卷积神经网络的图像分类算法研究[D].乌鲁木齐：新疆大学,2018.
[17]李晓雨,薛宏飞,高尚,等.基于改进残差网络的车型识别算法[J].信息技术与信息化,2020(02):121-126.
[18]申桐.基于卷积神经网络的多标记学习[D].北京：华北电力大学(北京),2021.
[19]韩冰.基于深度复卷积神经网络的滚动轴承故障诊断方法研究[D].北京：北京交通大学,2021.

相似文献/References:

[1]胡 茜,郑 望,张 娜.汽轮机虚拟调速器在冰机机组的应用[J].工业仪表与自动化装置,2015,(05):83.
　HU Qian,ZHENG Wang,ZHANG Na.Application of steam turbine virtual governor in ice machine unit[J].Industrial Instrumentation & Automation,2015,(04):83.

备注/Memo

备注/Memo:
收稿日期：2022-04-13
基金项目：国家科技支撑计划资助项目（2020BAA03B01）
作者简介：甘李（1985），男，工程师，主要从事电厂热工系统控制优化研究。
更新日期/Last Update: 1900-01-01