|本期目录/Table of Contents|

[1]李 闯,蔺奕存,李鹏竹,等.基于One-Class SVM的凝泵入口滤网堵塞预警模型开发与应用[J].工业仪表与自动化装置,2022,(05):97-102.[doi:10.19950/j.cnki.cn61-1121/th.2022.05.018]
　LI Chuang,LIN Yicun,LI Pengzhu,et al.Development and application of an early warning model for condensate pump inlet filter clogging based on One-Class SVM[J].Industrial Instrumentation & Automation,2022,(05):97-102.[doi:10.19950/j.cnki.cn61-1121/th.2022.05.018]
点击复制

基于One-Class SVM的凝泵入口滤网堵塞预警模型开发与应用

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

卷:
期数:
2022年05期
页码:
97-102
栏目:
出版日期:
2022-10-15

文章信息/Info

Title:
Development and application of an early warning model for condensate pump inlet filter clogging based on One-Class SVM
文章编号:
1000-0682（2022）05-0000-00
作者:
李 闯1; 蔺奕存2; 李鹏竹1; 谭祥帅2; 解世涛1; 吴青云2; 郭云飞2; 李 昭2; 姚 智2; 李雪冰1
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):
LI Chuang1;  LIN Yicun2;  LI Pengzhu1;  TAN Xiangshuai2;  XIE Shitao1;  WU Qingyun2;  GUO Yunfei2; LI Zhao2;  YAO Zhi2;  LI Xuebing1
(1.Jingneng Shiyan Thermal Power Co., Ltd., Shiyan ,442000, Hubei Shiyan,China2.Xi’an Thermal Power Research Institute Co., Ltd., Shaanxi Xi’an 710054, China)
关键词:
凝结水泵; 滤网堵塞; 机器学习; 异常检测; 预警模型
Keywords:
Condensate pump;  Filter sieve blocked;  Machine learning;  Anomaly detection;  Warning model
分类号:
TK229.2
DOI:
10.19950/j.cnki.cn61-1121/th.2022.05.018
文献标志码:
B
摘要:
电厂凝结水泵入口滤网在机组长周期运行时，由于凝结水水质变差，凝结水中杂质增多，易导致凝结泵入口滤网堵塞现象的发生。为解决以上问题，丰富在滤网堵塞初期对机组运行状态的监视手段，该研究采用基于单类支持向量机（One-Class SVM）算法，建立了凝结水泵入口滤网堵塞智能诊断与预警模型。通过提取与凝结水泵入口滤网堵塞具有因果联系的主要参数的历史数据，对智能预警模型进行训练优化，并完成了相关测试。研究结果表明，本模型可以有效对凝结水泵入口滤网发生堵塞现象进行识别，准确率达到99.96%，误报率低。通过试验测试数据对模型进行测试，结果表明该模型的召回率达到90.20%，准确率达到93.18%，满足工业生产需求，可有效指导机组操作人员及时采取相应措施，避免机组因非正常停机而造成经济损失。
Abstract:
During the long-term operation of the condensate pump inlet filter in the power plant, due to the deterioration of the condensate water quality and the increase of impurities in the condensate water, it is easy to cause the blockage of the condensate pump inlet filter. In order to solve the above problems and enrich the monitoring means of unit operation state in the early stage of filter blockage, an intelligent diagnosis and early warning model of condensate pump inlet filter blockage is established based on One-Class SVM algorithm. By extracting the historical data of the main parameters with causal relationship with the inlet filter plugging of the condensate pump, the intelligent early warning model is trained and optimized, and the related tests are completed. The results show that this model can effectively identify the clogging phenomenon of the inlet filter of the condensate pump, and the accuracy rate reaches 99.96 %, with low false alarm rate. The test results show that the recall rate of the model reaches 90.20 % and the accuracy rate reaches 93.18 %, which meets the needs of industrial production. It can effectively guide the operator of the unit to take corresponding measures in time to avoid economic losses caused by abnormal shutdown of the unit.

参考文献/References:

[1]安文凭,焦晓峰,曹磊,等.330 MW直接空冷机组凝结水泵出力不足原因分析与处理[J].内蒙古电力技术,2016,34(1):52-55.?div>[2]高耀岿,王林,高海东,等.火电厂智能控制系统体系架构及关键技术[J].热力发电,2022,51(3):166-174.

[3]贾志军,白德龙,宋燕杰,等.基于KPCA算法的电厂设备在线故障诊断研究[J].综合智慧能源,2021,43(8):48-53.
[4]明震,朱希安.基于LSTM网络的凝结水泵故障预警[J]. 北京信息科技大学学报:自然科学版,2021,36(6):53-57.
[5]曹顺安,胡家元,葛磊.基于数据挖掘及可信度理论的火电厂水汽化学过程故障诊断方法[J].计算机与应用化学, 2010(4):523-527.
[6]闻佳,王宏君,邓佳,等.基于深度学习的异常事件检测[J]. 电子学报,2020,48(2):308-313.?/div>
[7]ZONG B, SONG Q, MIN M R, et al. Deep autoencoding Gaussian mixture model for unsupervised anomaly detection[A]. Proceedings of the 6th International Conference on Learning Representations[C]. Vancouver，BC，Canada：OpenReview，2018. 1 - 19. [8]KUMARAGE T, RANATHUNGA S, KURUPPU C, et al. Generative adversarial networks (GAN) based anomaly detection in industrial software systems[C]//2019 Moratuwa Engineering Research Conference (MERCon). IEEE, 2019: 43-48.
[9]NHO Y H, RYU S, KWON D S. UI-GAN: Generative adversarial network-based anomaly detection using user initial information for wearable devices[J]. IEEE Sensors Journal, 2021, 21(8): 9949-9958.
[10]席亮,刘涵,樊好义,等.基于深度对抗学习潜在表示分布的异常检测模型[J].电子学报,2021,49(7):1257.
[11]武玉坤,李伟,倪敏雅,等.单类支持向量机融合深度自编码器的异常检测模型[J].计算机科学,2022,49(3):141-151.
[12]毛前军,梁致远,刘冬华,等.基于单类支持向量机的冷水机组温度传感器故障检测[J].制冷学报,2019,40(5):130-135.
[13]张玉彦.基于深度自编码器的机械故障诊断方法研究[D].武汉:华中科技大学,2019:3-5.?/div>
[14]刘建伟,谢浩杰,罗雄麟.生成对抗网络在各领域应用研究进展[J].自动化学报,2020:2500-2536.
[15]TAC D M J. One-class classification: Concept learning in the absence of counter-examples[J].2002.
[16]王璐. 660 MW 机组启动中凝结水系统故障及处理[J]. 华电技术,2018,40(4):52-53.
[17]PADMAJA T M, LAKSHMI P J. A hybrid rule extraction method for one-class support vector machines[C]//2015 IEEE 9th International Conference on Intelligent Systems and Control (ISCO). IEEE, 2015: 1-4.
[18]张士强,戴计生,徐海龙,等.一种基于循环神经网络的牵引电机滤网堵塞预警方法[J].控制与信息技术,2021.
[19]宗全利,杨洪飞,刘贞姬,等.网式过滤器滤网堵塞成因分析与压降计算[J].农业机械学报,2017,48(9):215-222.
[20]BARBADO A, CORCHO , BENJAMINS R. Rule extraction in unsupervised anomaly detection for model explainability: Application to OneClass SVM[J]. Expert Systems with Applications, 2022, 189: 116100.
[21]PADMAJA T M, LAKSHMI P J. A hybrid rule extraction method for one-class support vector machines[C]//2015 IEEE 9th International Conference on Intelligent Systems and Control (ISCO). IEEE, 2015: 1-4.

相似文献/References:

备注/Memo

备注/Memo:
收稿日期：2022-06-10

基金项目：
京能集团科技项目（JT202027）

作者简介：
李闯（1987），河南新乡人，本科，工程师，从事发电厂智能控制研究及热工控制系统优化。
更新日期/Last Update: 1900-01-01