小波奇异性检测诊断有载分接开关故障

Wavelet Singularity Detection Used for Fault Diagnosis of On-load Tap Changers

为有效判断出有载分接开关(OLTC)的状况,测量了OLTC切换过程中其表面的振动信号,分析了切换动作过程。运用Hilbert变换提取了振动包络,采用小波变换及模极大值算法滤除了包络中的高频成分及随机噪声,结果表明该算法能够抑制OLTC振动包络信号中的高频及随机噪声。提出了OLTC振动包络信号各波峰的奇异性指数的计算方法,奇异性指数的物理意义就是信号在奇异点处的光滑程度,或者说是信号峰值陡峭程度的一个度量,所以包络信号各波峰的奇异性指数可用来判别OLTC的动静触头表面状况、弹簧压紧情况等。因此,运用奇异性指数作为特征参数,分析了触头正常及触头表面磨损、触头表面烧损、触头弹簧松动时的OLTC切换时表面产生的振动信号,提出与触头正常时相比,触头磨损及触头弹簧松动时其相应的振动包络信号波峰的奇异性指数变大;触头烧损时其相应的振动包络信号波峰的奇异性指数变小,结果表明这种基于包络分析及小波奇异性的算法可用于振动法检测OLTC状况。

For reducing on-load tap-changer （OLTC） failures and assuring the saferunning of on-load regulation transformer, it is important to develop the on-linemonitoring technology of OLTCs. The operation of an OLTC on a high-voltage powertransformer produces a vibration signal that consists of a series of well-definedtransients. The singularity exponent which marks the cragged ness degree of wavecrest can be used to analysis the OLTC vibration envelope and it is an efficient method to diagnose the condition of OLTC. In this paper, the vibration signals of OLTC surface are measured and the OLTC change process is analyzed. The envelope of vibration signal produced by the OLTC switching operation is extracted by the Hilbert transform. The high frequency components and random noise of envelope was depressed by wavelet transform and modulus maximum algorithm, and the experimental results in the lab show that above algorithm is very effective to depress the high frequency and random noise contained in OLTC vibration envelope signals. The method used to calculate the singularity exponent of each wave crest is presented. It is known that the physical meaning of singularity exponent is the smoothness degree of a signal at singularity point, in the other words, it indicates the craggedness degree of the signal＇s crest. That is to say that the singu- larity exponent can be used to assess the surface condition of OLTCs moving and fixed contacts, the compaction condition of spring and etc. In this paper using the singularity exponent as characteristic parameter, this paper analyzes the OLTC surface Vibration envelope when the OLTC contact is normal, worn, burn-out and spring-loosing respectively. Compared with the normal contacts, the corresponding wave crest＇s singularity exponents of worn and springloosing contact become larger and burn-out contact become smaller. The results indicate that this algorithm based on envelope analysis and wavelet singularity can be used to diagnose the condition of OLTC.