发电技术 ›› 2022, Vol. 43 ›› Issue (6): 951-958.DOI: 10.12096/j.2096-4528.pgt.22162

• 发电及环境保护 • 上一篇    下一篇

基于概率神经网络-小波神经网络-DS信息融合的电厂引风机故障诊断

张航1, 周传杰1, 张林1, 陈节涛1, 徐春梅2, 彭道刚2   

  1. 1.国电长源汉川第一发电有限公司,湖北省 武汉市 431614
    2.上海电力大学自动化工程学院,上海市 杨浦区 200090
  • 收稿日期:2022-10-24 出版日期:2022-12-31 发布日期:2023-01-03
  • 作者简介:张航(1987),男,工程师,主要从事火电厂热控设备检修维护、DCS自动控制系统日常维护、优化工作,543134425@qq.com
    周传杰(1973),男,工程师,主要从事电厂热工自动控制研究工作,51623588@qq.com
    张林(1978),男,工程师,主要从事火电厂炉侧热控设备检修维护的技术管理工作,zlinz6666@sina.com
    陈节涛(1989),男,硕士,工程师,主要从事火电厂热控设备检修维护、DCS自动控制系统日常维护、优化工作,taoshengyijiuchen@163.com
    徐春梅(1982),女,博士,讲师,研究方向故障诊断与状态评价等,本文通信作者,chunmeixu@yeah.net
    彭道刚(1977),男,博士,教授,主要研究方向为低碳智能发电、综合智慧能源与能源互联网等,pengdaogang@126.com
  • 基金资助:
    上海市“科技创新行动计划”高新技术领域项目(22511103800);国电长源电力股份有限公司科技项目(HCYF-SCFW-2021-127)

Fault Diagnosis of Power Plant Induced Draft Fan Based on PNN-WNN-DS Information Fusion

Hang ZHANG1, Chuanjie ZHOU1, Lin ZHANG1, Jietao CHEN1, Chunmei XU2, Daogang PENG2   

  1. 1.Guodian Changyuan Hanchuan No. 1 Power Generation Co. , Ltd. , Wuhan 431614, Hubei Province, China
    2.College of Automation Engineering, Shanghai University of Electric Power, Yangpu District, Shanghai 200090, China
  • Received:2022-10-24 Published:2022-12-31 Online:2023-01-03
  • Supported by:
    Shanghai Science and Technology Commission Program(22511103800);CHN Energy Changyuan Electric Power Co., Ltd(HCYF-SCFW-2021-127)

摘要:

针对电厂引风机工况复杂、工作环境恶劣、易出现故障等问题,提出了一种基于改进D-S证据理论的融合诊断方法。该方法利用概率神经网络(probabilistic neural network,PNN)和小波神经网络(wavelet neural network,WNN)对测试样本进行初步诊断,并形成证据体,再利用改进D-S融合方法进行融合诊断。该融合方法根据证据体的信任度和焦元的信任度分配冲突信息,使得信任度高的焦元支持率得到加强、信任度低的焦元支持率得到削弱,融合结果更为合理。仿真结果表明,融合故障诊断方法能有效地避免误诊现象,提高了诊断的正确率,且能合理分配冲突信息。

关键词: 电厂引风机, 焦元, 故障诊断, 改进D-S证据理论

Abstract:

Aiming at the problems of complex operating conditions of induced draft fan, harsh working environment, and easy failure of power plant induced draft fan, a fault diagnosis method of the improved dempster-shafer evidence theory was proposed. In this method, the probabilistic neural network (PNN) and wavelet neural network (WNN) were used for preliminary diagnosis, and the evidence bodies were formed according to the output of PNN and WNN. Then the improved D-S fusion method was used for fusion diagnosis. The improved D-S method distributes conflict information according to the trust degree of the evidence and the focal element, so that the support rate of the focal element with high trust degree is strengthened, and the focal element with low trust degree is weakened, which makes the fusion diagnosis result more reasonable. The simulation results show that the proposed method can effectively diagnose the vibration fault of induced draft fan, avoid misdiagnosis, improve the accuracy of diagnosis, and reasonably distribute conflicting information.

Key words: power plant induced draft fan, focal element, fault diagnosis, improved D-S evidential theory

中图分类号: