Method for extracting characteristic quantity of vibration signal of transformer

Abstract

The invention discloses a method for extracting the characteristic quantity of a vibration signal of a transformer. According to the method, filter parameters are calculated based on a fast spectral kurtosis diagram method and a filter is designed to filter an original vibration signal; on the basis of an improved empirical mode decomposition method, the signal is decomposed to obtain a series ofintrinsic mode functions, correlation coefficients between all intrinsic mode function components and the original signal are calculated according to a correlation coefficient method, and a sensitiveintrinsic mode function is extracted and is reconstructed; and then the reconstructed transformer vibration signal characteristic frequency, kurtosis, amplitude and mean are extracted to form a vibration signal characteristic vector and the vibration signal characteristic vector is compared with the normal characteristic vector to determine the state of the transformer.

Description

technical field [0001] The invention relates to the field of transformer fault detection, in particular to a method for extracting characteristic quantities of transformer vibration signals based on a fast spectral kurtosis map method and an improved set empirical mode decomposition and correlation coefficient method. Background technique [0002] The research on diagnostic technology based on transformer vibration signals is mainly divided into two directions: simulation modeling and experimental research, and signal processing. In terms of signal processing, Fourier transform is the basis of signal frequency domain analysis, which reflects the distribution of the global frequency of the signal. Some researchers have proposed that the ratio of the fundamental frequency of the transformer vibration signal and the ratio of high-order harmonics should be used as the basis for transformer state diagnosis. Feature amount. With the further development of signal processing techno...

AI Technical Summary

Problems solved by technology

The time spectrum is mainly to monitor the sudden change of the signal on the time axis. The change of the transformer state is mostly the process of long-term accumulation of abnormal components. During this process, the signal is still an irregular periodic signal, and the frequency distribution of the vibration signal does not change with time. Time-spectrum analysis still has some limitations

In addition, the existing transformer vibration feature extraction technology seldom considers the denoising of the signal

Images