摘要
有载分接开关(OLTC)是变压器中唯一频繁动作的关键部件,容易发生机械故障,从而引发严重的安全事故。分析了V型OLTC的切换原理,设置了3种常见机械故障,利用多通道振动传感器采集切换过程的振动信号,提出了一种基于振动信号分析的V型OLTC机械故障识别方法。首先,采用CCEMDAN算法分解振动信号,基于相关系数法构建特征矩阵,利用奇异值分解对振动信号进行特征提取;其次,通过主成分分析与K-means聚类进行振动信号的最优测点选择;最后,基于XGBoost集成学习算法实现了机械故障分类。试验结果表明,OLTC箱体正面所测振动信号能更好地反映其内部机械状态,所提方法对于3种机械故障的识别准确率达到98%,高于随机森林、GBDT等集成学习算法。
The on-load tap changer(OLTC)is a critical component in transformers,which undergoes frequent operations and is prone to mechanical faults,leading to serious safety incidents.The switching principles of the V-type OLTC is analyzed,and three common mechanical defects are introduced on a specific V-type OLTC.Multi-channel vibration sensors are employed to collect vibration signals during the switching processes.A vibration signal analysis-based method for mechanical fault identification in V-type OLTCs is proposed.Firstly,the CCEMDAN algorithm is used to decompose the vibration signals.A feature matrix is constructed based on the correlation coefficient method,and singular value decomposition(SVD)is utilized for feature extraction.Secondly,optimal measurement points for vibration signals are selected through principal component analysis and Kmeans clustering.Finally,mechanical fault classification is achieved using the XGBoost ensemble learning algorithm.Experimental results demonstrate that the vibration signals measured on the front side of the OLTC enclosure better reflect its internal mechanical state.The proposed method achieves an identification accuracy of 98%for the three types of mechanical faults,surpassing ensemble learning algorithms such as random forests and GBDT.
作者
刘浩宇
高树国
邢超
王丽丽
郭小凡
LIU Haoyu;GAO Shuguo;XING Chao;WANG Lili;GUO Xiaofan(State Grid Hebei Electric Power Research Institute,Shijiazhuang 050021,China)
出处
《河北电力技术》
2024年第2期57-63,共7页
Hebei Electric Power
基金
国网河北省电力有限公司科技项目(kjcb2022-039,B704DY220110)。
关键词
有载分接开关
机械故障
振动信号
特征提取
集成学习
on-load tap changer
mechanical fault
vibration signal
feature extraction
ensemble learning
作者简介
刘浩宇(1993-),男,工程师,主要从事变压器状态感知与智能评估技术研究工作。
