基于时移多尺度注意熵和随机森林的水电机组故障诊断  

Fault diagnosis of hydropower units based on time-shifted multiscale attention entropy and random forest

作　　者：陈飞[1] 王斌[1] 刘婷[1] 张文静[1] 高园晨 陈帝伊[1] CHEN Fei;WANG Bin;LIU Ting;ZHANG Wenjing;GAO Yuanchen;CHEN Diyi(College of Water Resources and Architectural Engineering,Northwest A&F University,Yangling 712100,China)

机构地区：[1]西北农林科技大学水利与工程建筑学院,陕西杨凌712100

出　　处：《水利学报》2022年第3期358-368,378,共12页Journal of Hydraulic Engineering

基　　金：国家自然科学基金项目(51509210);陕西省重点研发计划项目(2021NY-181)。

摘　　要：针对传统诊断方法难以在高噪声环境下进行故障精准识别的问题,本文提出了一种抗噪性能良好、识别率高的水电机组故障诊断方法。首先,基于分形理论,提出了一种度量信号复杂度的工具——时移多尺度注意熵(Time-shifted multiscale attention entropy,TSMATE)。然后,利用主成分分析(Principal component analysis,PCA)对TSMATE进行降维处理,克服了特征冗杂问题。最终,将降维后的特征输入到随机森林(Random forests,RF)模型进行诊断。通过对振动信号添加不同信噪比的噪声,探究不同噪声强度下所提模型的抗噪性能。仿真实验表明,TSMATE-PCA-RF在0 dB、1 dB、2 dB以及3 dB四种不同信噪比噪声干扰下,分别取得了98.06%、98.89%、99.17%以及99.17%的诊断率,验证了所提模型具有良好的抗噪性能。该研究为水电机组故障诊断提供了新手段。In view of the fact that traditional diagnosis methods can not accurately identify faults in high noise environment,a fault diagnosis method for hydropower units with good anti-disturbance performance and high recognition rate is proposed in this paper.First,based on the fractal theory,a time-shifted multiscale attention entropy(TSMATE)is proposed to measure signal complexity.Second,principal component analysis(PCA)is used to reduce the dimension of TSMATE and overcome the problem of feature redundancy.Finally,the reduced dimension features are input into the random forests(RF)model for diagnosis.By adding noise with different signal-to-noise ratio to the vibration signals,the anti-noise performance of the proposed model under different noise intensity is investigated.The simulation results show that TSMATE-PCA-RF achieves diagnosis rates of 98.06%,98.89%,99.17%and 99.17%under the interference of four different signal-to-noise ratios of 0 dB,1 dB,2 dB and 3 dB,which verifies the good anti-noise performance of the proposed model.This study provides a new means for fault diagnosis of hydropower generating units.

关 键 词：时移多尺度注意熵 随机森林 主成分分析 水电机组 故障诊断 

分 类 号：TM312[电气工程—电机]