摘要
Abstract
The voltage damping oscillation method is effective for diagnosing winding faults,but it still has the following problems.The voltage-damping oscillation of transformer winding continues to fluctuate slightly for a period before the end of charging and discharging.In this process,there are a lot of irregular maximum points,minimum points,and burrs,which interfere with the subsequent characteristic processing and oscillation wave analysis.According to the existing literature,the oscillation ends when it attenuates to 5%of the maximum value,and all poles in this band are effective extreme points.Through a large number of experiments,it is observed that when the voltage-damping oscillation is no longer attenuated,its stable value often exceeds this threshold.As a result,the definition of the effective wave band of the above voltage-damping oscillation is inapplicable.Due to the complexity of on-site testing and external interference,two voltage-damping oscillation signals measured at different time intervals on the same transformer may have small deviations.However,the time-domain feature extraction of the oscillation wave is distinguished only by a single curve feature,and the identifiable accuracy is low when diagnosing the fault location and degree.
Therefore,this paper presents a transformer winding fault analysis method for selecting a dynamic wave band of voltage-damping oscillations.First,a transformer winding fault simulation platform is built,considering four kinds of faults:axial displacement,inter-disk capacitance,inter-pie,and inter-turn short circuits.Secondly,the characteristics of oscillation wave attenuation are analyzed from the perspective of energy conversion,selecting the effective wave band through a defined energy attenuation factor.Then,the waveform feature correlation degree(FCD)is obtained for fault type identification.At the same time,considering the rich feature information between the extreme point and the waveform,the oscillating wave binary image is constructed based on mathematical morphology to eliminate interference and extract more stable Tamura texture features.Finally,according to the distribution rule of characteristic parameter combination,the application analysis is carried out through the actual transformer.
The dynamically selected band shows minimal interference,clear attenuation oscillation regularity,and rich feature information.Oscillation waves under four fault types are different from each other,exhibiting significant variation patterns compared to normal windings.The waveform FCD is effective in identifying winding fault types.Oscillation waves exhibit minor differences in fault areas and degrees under the same fault type.Different combinations of four Tamura texture features demonstrate good performance in classifying fault degrees and types.In general,the waveform FCD and Tamura texture feature based on binarization values extracted under different faults have apparent separation and clustering,which can recognize fault type,degree,and region.关键词
变压器/振荡波/绕组故障/动态波段/波形特征/纹理特征Key words
Transformer/voltage damping oscillation/winding fault/dynamic band/waveform feature/texture feature分类
信息技术与安全科学
