智能电网
智能電網
지능전망
Smart Grid
2014年
8期
8-12
,共5页
蒋丹%赵书涛%赵现平%任剑峰
蔣丹%趙書濤%趙現平%任劍峰
장단%조서도%조현평%임검봉
高压断路器%盲源分离%三维图%极限学习机%故障诊断
高壓斷路器%盲源分離%三維圖%極限學習機%故障診斷
고압단로기%맹원분리%삼유도%겁한학습궤%고장진단
high voltage circuit breakers%blind source separation%three-dimensional diagram%extreme learning machine (ELM)%fault diagnosis
为了完善现有断路器故障诊断方法,针对其当前存在的不足,提出极限学习机(extreme learning machine,ELM)与振声时三维故障诊断相结合的高压断路器故障诊断方法,并给出可行的诊断步骤和分析方法。此方法首先利用盲源分离技术将混杂在断路器声波信号中的汽车鸣笛、打雷等强干扰噪声剔除,然后利用多信号时标对准的方法构建振声时三维图,提取断路器振声时三维图的包络,最后采用基于形状的分级特征向量提取方法构建极限学习机的输入向量进行故障诊断。其中,一级特征提取振声时三维图包络六面八角的特征向量,二级特征提取振声时三维图二维投影的区域弦的分布特征。实验结果表明,该方法可以用于诊断高压断路器机械故障。
為瞭完善現有斷路器故障診斷方法,針對其噹前存在的不足,提齣極限學習機(extreme learning machine,ELM)與振聲時三維故障診斷相結閤的高壓斷路器故障診斷方法,併給齣可行的診斷步驟和分析方法。此方法首先利用盲源分離技術將混雜在斷路器聲波信號中的汽車鳴笛、打雷等彊榦擾譟聲剔除,然後利用多信號時標對準的方法構建振聲時三維圖,提取斷路器振聲時三維圖的包絡,最後採用基于形狀的分級特徵嚮量提取方法構建極限學習機的輸入嚮量進行故障診斷。其中,一級特徵提取振聲時三維圖包絡六麵八角的特徵嚮量,二級特徵提取振聲時三維圖二維投影的區域絃的分佈特徵。實驗結果錶明,該方法可以用于診斷高壓斷路器機械故障。
위료완선현유단로기고장진단방법,침대기당전존재적불족,제출겁한학습궤(extreme learning machine,ELM)여진성시삼유고장진단상결합적고압단로기고장진단방법,병급출가행적진단보취화분석방법。차방법수선이용맹원분리기술장혼잡재단로기성파신호중적기차명적、타뢰등강간우조성척제,연후이용다신호시표대준적방법구건진성시삼유도,제취단로기진성시삼유도적포락,최후채용기우형상적분급특정향량제취방법구건겁한학습궤적수입향량진행고장진단。기중,일급특정제취진성시삼유도포락륙면팔각적특정향량,이급특정제취진성시삼유도이유투영적구역현적분포특정。실험결과표명,해방법가이용우진단고압단로기궤계고장。
In order to improve the method of the existing circuit breaker fault diagnosis, based on the existing shortcomings of mechanical fault diagnosis, a new method for fault diagnosis is proposed based on three-dimensional diagram of vibration, acoustic and time and ELM, and feasible diagnostic steps and analysis are also introduced. It uses the method of blind source separation to remove car horns, thunder and other strong interference noise in acoustic signals of the circuit breaker, and then builds three-dimensional diagram of vibration, acoustic and time based on multiple signal time-alignment, and then extracts the envelope of the three-dimensional diagram. At last, an extraction method based on shape classification feature vector is used to build the input vector of ELM to conduct fault diagnosis. Among them, the first level of feature extracts the three-dimensional diagram envelope’s six sides and eight angle characteristic vector; the secondary feature extracts the features of region chordal distribution of two-dimensional projection of three-dimensional diagram. The experiment shows that the proposed method is effective to diagnose the mechanical faults of high voltage circuit breakers.