[1]许凯,富威,陈世均,等.基于LSTM神经网络的汽轮发电机状态监测系统[J].应用科技,2020,47(6):96-100.[doi:10.11991/yykj.202007003]
　XU Kai,FU Wei,CHEN Shijun,et al.State monitoring system of turbine generator based on LSTM neural network[J].Applied science and technology,2020,47(6):96-100.[doi:10.11991/yykj.202007003]
点击复制

基于LSTM神经网络的汽轮发电机状态监测系统(/HTML)
分享到：

《应用科技》[ISSN:1009-671X/CN:23-1191/U]

卷:
第47卷
期数:
2020年6期
页码:
96-100
栏目:
机电工程
出版日期:
2021-01-31

文章信息/Info

Title:
State monitoring system of turbine generator based on LSTM neural network
作者:
许凯1;  富威1;  陈世均2;  孟宇龙3;  马佳瑞1
1. 哈尔滨工程大学 机电工程学院，黑龙江 哈尔滨 150001;
2. 中广核苏州热工研究院有限公司，江苏 苏州 215000;
3. 哈尔滨工程大学 计算机科学与技术学院，黑龙江 哈尔滨 150001
Author(s):
XU Kai1;  FU Wei1;  CHEN Shijun2;  MENG Yulong3;  MA Jiarui1
1. College of Mechanical and Electrical Engineering, Harbin Engineering University, Harbin 150001, China;
2. Suzhou Nuclear Power Research Institute, China General Nuclear Power Corporation, Suzhou 215000, China;
3. College of Computer Science and Technology, Harbin Engineering University, Harbin 150001, China
关键词:
机器学习; 健康预测; 汽轮发电机; 数据处理; 预测性维护; LSTM神经网络; 状态监测系统; 信号分析
Keywords:
machine learning; health prediction; turbo-generator; data processing; preventive maintenance; LSTM neural network; condition monitoring system; signal analysis
分类号:
TM623.3
DOI:
10.11991/yykj.202007003
文献标志码:
A
摘要:
为了快速获取汽轮发电机运行参数及其分析处理结果，完成运行状态的判断识别，本文以某核电厂使用的汽轮发电机状态参数为分析依据进行监测研究。先分别以机器学习算法中的线性回归和长短期记忆(LSTM)神经网络为数据处理方法，对2种方法得到的状态参数预测结果进行对比分析，确定以LSTM神经网络为状态监测核心算法，设计并搭建了汽轮发电机状态监测系统。之后采用传统的振动信号分析方法对汽轮发电机运行状态监测结果进行验证分析，保证监测结果准确可靠。实验结果表明：基于LSTM神经网络的状态监测方法能够有效处理运行参数，提取数据中隐含的设备状态信息，实现汽轮发电机状态实时监测，提高发电机组整体工作的稳定性与安全性，降低故障导致机组停机损毁的概率。
Abstract:
In order to quickly obtain the operating parameters of turbo-generator and the results of analysis and processing, and complete the judgment and identification of the operating state, the monitoring and research is carried out on the basis of the analysis of the turbo-generator state parameters used in a nuclear power plant. Firstly, the linear regression and long short term memory (LSTM) neural network in machine learning algorithm are used as data processing methods, and the prediction results of state parameters obtained by the two methods are compared and analyzed. Based on the LSTM neural network, the condition monitoring system of turbo-generator is designed and built. Finally, the traditional vibration signal analysis method is used to verify and analyze the monitoring results of the running state of turbo-generator, ensuring the accuracy and reliability of the monitoring results. The results showed that the condition monitoring method based on LSTM neural network can deal with the operation parameters effectively, extract the equipment state information implied in the data, realize the real-time monitoring of the turbo-generator condition, improve the stability and safety of the whole operation of the generator set, and reduce the probability of the unit shutdown damage caused by the failure.

参考文献/References:

[1] 张雁鹏, 姜茜. 汽轮发电机重大事故举隅[J]. 东方电气评论, 2015, 29(1): 65-72
[2] RAHEJA D, LLINAS J, NAGI R, et al. Data fusion/data mining-based architecture for condition-based maintenance[J]. International journal of production research, 2006, 44(14): 2869-2887.
[3] BASIR O, YUAN X H. Engine fault diagnosis based on multi-sensor information fusion using Dempster-Shafer evidence theory[J]. Information fusion, 2007, 8(4): 379-386.
[4] 董文婷. 基于大数据分析的风电机组健康状态的智能评估及诊断[D]. 上海: 东华大学, 2016.
[5] SON J, KANG D, BOO D, et al. An experimental study on the fault diagnosis of wind turbines through a condition monitoring system[J]. Journal of mechanical science and technology, 2018, 32(12): 5573-5582.
[6] XIE Lei, ZENG Jiusun, KRUGER U, et al. Fault detection in dynamic systems using the Kullback-Leibler divergence[J]. Control engineering practice, 2015, 43: 39-48.
[7] JIAO Weidong, QIAN Suxiang, CHANG Yongping, et al. Research on vibration response of a multi-faulted rotor system using LMD-based time-frequency representation[J]. EURASIP journal on advances in signal processing, 2012, 2012(1): 73.
[8] 张德利. 基于贝叶斯网络的故障智能诊断方法研究[D]. 保定: 华北电力大学(河北), 2008.
[9] 黄乃成, 顾煜炯, 谢骐宇, 等. 基于支持向量机与知识的汽轮发电机组智能故障诊断研究[J]. 现代电力, 2012, 29(11): 68-73
[10] 程军圣, 郑近德, 杨宇. 一种新的非平稳信号分析方法——局部特征尺度分解法[J]. 振动工程学报, 2012, 25(2): 215-220
[11] 杨宇, 李永国, 何知义, 等. 基于LCD的齿轮箱混合故障盲源分离研究[J]. 中国机械工程, 2015, 26(8): 1062-1066
[12] 仝蕊, 康建设, 孙健, 等. 基于局部特征尺度分解与复合谱分析的齿轮性能退化特征提取[J]. 兵工学报, 2019, 40(5): 1093-1102
[13] 刘海兰, 李小平, 芮延年. 基于时域平均和Hilbert-Huang变换的时频熵理论轧机齿轮箱故障诊断[J]. 机械传动, 2011, 35(9): 54-57
[14] 李敏通, 杨青, 宋蒙, 等. 多特性参数相结合的柴油机故障振动信号特征提取与诊断[J]. 汽车工程, 2014, 36(4): 438-442
[15] 雷亚国, 贾峰, 周昕, 等. 基于深度学习理论的机械装备大数据健康监测方法[J]. 机械工程学报, 2015, 51(21): 49-56
[16] 雷亚国, 杨彬, 杜兆钧, 等. 大数据下机械装备故障的深度迁移诊断方法[J]. 机械工程学报, 2019, 55(7): 1-8
[17] 黄郑, 王红星, 于海泉, 等. 基于多模型鲁棒输入训练神经网络协同的燃气-蒸汽联合循环机组传感器故障诊断方法[J]. 中国电力, 2019, 52(11): 125-133
[18] POUR F K, PUIG V, CEMBRANO G. Health-aware LPV-MPC based on a reliability-based remaining useful life assessment[J]. IFAC-PapersOnLine, 2018, 51(24): 1285-1291.
[19] BAUR M, ALBERTELLI P, MONNO M. A review of prognostics and health management of machine tools[J]. The international journal of advanced manufacturing technology, 2020, 107(5): 2843-2863.
[20] 潘作为. 大功率电站风机振动监测及故障诊断方法研究[D]. 北京: 华北电力大学(北京), 2016.
[21] ALZGHOUL A, L?FSTRAND M, BACKE B. Data stream forecasting for system fault prediction[J]. Computers & industrial engineering, 2012, 62(4): 972-978.
[22] 胡海峰, 安茂春, 秦国军, 等. 基于隐半Markov模型的故障诊断和故障预测方法研究[J]. 兵工学报, 2009, 30(1): 69-75
[23] 杜乐. 基于状态监测数据的盘式刀库故障预测与健康管理技术研究[D]. 长春: 吉林大学, 2018.
[24] 柯赟, 宋恩哲, 姚崇, 等. 船舶柴油机故障预测与健康管理技术综述[J]. 哈尔滨工程大学学报, 2020, 41(1): 125-131

相似文献/References:

[1]王立国,杜心平.K均值聚类和孪生支持向量机相结合的高光谱图像半监督分类[J].应用科技,2017,44(03):12.[doi:10.11991/yykj.201606010]
　WANG Liguo,DU Xinping.Semi-supervised classification of hyperspectral images applying the combination of K-mean clustering and twin support vector machine[J].Applied science and technology,2017,44(6):12.[doi:10.11991/yykj.201606010]
[2]田瑶瑶,张惠娟,杨忠,等.基于小波包和SOM神经网络的电作动器故障诊断[J].应用科技,2018,45(01):1.[doi:10.11991/yykj.201706023]
　TIAN Yaoyao,ZHANG Huijuan,YANG Zhong,et al.Fault diagnosis of electromechanical actuator based on wavelet packet and SOM neural network[J].Applied science and technology,2018,45(6):1.[doi:10.11991/yykj.201706023]
[3]王信,汪友生.基于深度学习与传统机器学习的人脸表情识别综述[J].应用科技,2018,45(01):65.[doi:10.11991/yykj.201707008]
　WANG Xin,WANG Yousheng.Facial expression recognition based on deep learning and traditional machine learning[J].Applied science and technology,2018,45(6):65.[doi:10.11991/yykj.201707008]
[4]程子一,刘志林.改进的核相关滤波算法在自航模动态目标跟踪应用[J].应用科技,2019,46(01):36.[doi:10.11991/yykj.201805009]
　CHENG Ziyi,LIU Zhilin.Application of improved kernel correlation filtering algorithm in small ship dynamic target tracking[J].Applied science and technology,2019,46(6):36.[doi:10.11991/yykj.201805009]
[5]田洪晨,王立国,赵亮,等.结合波段选择的差分进化高光谱图像分类[J].应用科技,2019,46(05):45.[doi:10.11991/yykj.201811019]
　TIAN Hongchen,WANG Liguo,ZHAO Liang,et al.Differential evolution hyperspectral image classification combined with band selection[J].Applied science and technology,2019,46(6):45.[doi:10.11991/yykj.201811019]
[6]胡强,屈蔷,何鑫.改进的多特征融合人行道检测算法[J].应用科技,2020,47(2):35.[doi:10.11991/yykj.201906016]
　HU Qiang,QU Qiang,HE Xin.An improved sidewalk detection algorithm based on multi-feature fusion[J].Applied science and technology,2020,47(6):35.[doi:10.11991/yykj.201906016]
[7]余移山,陆继东,董美蓉,等.T91金属管道表面特性对LIBS测量的影响[J].应用科技,2020,47(4):82.[doi:10.11991/yykj.202001002]
　YU Yishan,LU Jidong,DONG Meirong,et al.Study on the influence of surface characteristics of T91 metal pipeline on laser-induced breakdown spectroscopy measurement[J].Applied science and technology,2020,47(6):82.[doi:10.11991/yykj.202001002]

备注/Memo

备注/Memo:
收稿日期:2020-07-03。
作者简介:许凯，男，硕士研究生;富威，男，副教授，博士
通讯作者:富威，E-mail：fuwei@hrbeu.edu.cn
更新日期/Last Update: 2021-02-05