基于SVMD-CMSEE的油侵变压器PD高频脉冲IMF异常峰值点检测

摘要:

油浸变压器PD高频脉冲信号由非单一振动源所产生的复杂信号连续叠加而成，在对高频脉冲信号进行分解时，会出现模态混叠问题，即不同模态分量之间相互干扰，难以准确分离出每个模态分量的特征，进而难以设定合理的阈值，进而产生异常峰值点检测精度低的问题。对此，设计基于SVMD-CMSEE的油侵变压器PD高频脉冲IMF异常峰值点检测算法。针对高频脉冲信号实施逐次变分模态分解（SVMD），计算其复合多尺度能量熵（CMSEE），进而选取有效的IMF分量，实现基于SVMD-CMSEE的脉冲信号IMF分量取样。根据脉冲信号的占空比，确定PD条件下的瞬时电压与电流，以求解变压器高频脉冲强度，分析局部放电下的油侵变压器高频脉冲强度。对去噪后的高频脉冲信号，实施IMF分解处理，在混叠的高频脉冲信号中分离出每个模态分量的特征，再结合有效的IMF分量设定检测阈值，从而识别异常峰值点，完成对PD高频脉冲IMF异常峰值点的检测。实验结果表明，存在局部放电问题时，基于上述方法能够有效分辨出单一波段内的高频脉冲成分，且在网侧绕组、阀侧绕组注入脉冲的情况下，都可以准确检测出异常的脉冲峰值，有助于实现对油侵变压器异常放电信号的识别。

Abstract:

The high-frequency pulse signal of oil immersed transformer PD is composed of complex signals generated by non single vibration sources that are continuously superimposed. When decomposing the high-frequency pulse signal, mode mixing problems may occur, that is, different modal components interfere with each other, making it difficult to accurately separate the characteristics of each modal component, and thus difficult to set a reasonable threshold, resulting in low accuracy in detecting abnormal peak points. Design an algorithm for detecting abnormal peak points of PD high-frequency pulse IMF in oil invasion transformers based on SVMD-CMSEE. Implementing successive variational mode decomposition (SVMD) for high-frequency pulse signals, calculating their composite multiscale energy entropy (CMSEE), and selecting effective IMF components to achieve pulse signal IMF component sampling based on SVMD-CMSEE. Determine the instantaneous voltage and current under PD conditions based on the duty cycle of the pulse signal, in order to solve the high-frequency pulse intensity of the transformer and analyze the high-frequency pulse intensity of oil invasion transformer under partial discharge. Perform IMF decomposition on the denoised high-frequency pulse signal, separate the characteristics of each modal component in the mixed high-frequency pulse signal, and then set the detection threshold based on the effective IMF component to identify abnormal peak points and complete the detection of PD high-frequency pulse IMF abnormal peak points. The experimental results show that when there is a partial discharge problem, the above method can effectively distinguish the high-frequency pulse components within a single wavelength band, and can accurately detect abnormal pulse peaks in the case of injecting pulses into the grid side winding and valve side winding, which is helpful for identifying abnormal discharge signals of oil invasion transformers.