Extraction method of OLTC impact characteristic SVD denoised based on S-transformed time-frequency spectrum

Abstract

The invention discloses an extraction method of an OLTC impact characteristic SVD denoised based on an S-transformed time-frequency spectrum. The extraction method comprises the following steps of: (1) collecting a vibration signal of an on-load tap-changer OLTC in a normal state by an acceleration sensor; (2) performing an S-transformation on a collected vibration signal, and performing discretization to obtain an S-transformed time-frequency spectrum; (3) obtaining a spectral coefficient matrix A based on the time-frequency spectrum, and performing singular value decomposition on the matrixA to obtain each singular value of the matrix A; (4) setting a threshold [sigma]th, and setting a singular value less than or equal to the [sigma]th in a singular value sequence [sigma] to zero, and then reconstructing an S-transformed time spectral coefficient matrix B; and (5) performing an inverse S-transformation on the time spectral coefficient matrix B to obtain a time-domain impact characteristic of the vibration signal. The extraction method achieves a good effect on extracting the impact characteristic of the vibration signal of the tap-changer.

Description

technical field [0001] The invention relates to the technical field of on-load tap changer fault diagnosis, in particular to an OLTC vibration signal processing method based on S-transform SVD noise reduction. Background technique [0002] As one of the core components of the transformer, the transformer on-load tap changer OLTC plays an important role in stabilizing the load center voltage, adjusting the reactive power flow, and increasing the flexibility of power grid dispatching in the power system. During the operation of the on-load tap-changer, the collision or friction between the mechanism parts will generate vibration signals, which contain rich equipment status information. However, the vibration signals collected on site often contain abnormal data and various noises, which will affect the results of vibration signal analysis. Therefore, before analyzing the vibration signal, choosing an appropriate method to extract the impulse response at the moment of OLTC swi...

AI Technical Summary

Problems solved by technology

The high-order statistics method uses an optimized objective function to construct a blind deconvolution filter to detect weak impact components in the signal, but the length of the filter in this method is difficult to determine, which limits its application

The kurtosis method uses the weighted kurtosis index combined with the kurtosis index and the cross-correlation coefficient as the optimization target, and it is feasible to extract the signal impact component by using the stochastic resonance detection method. However, for the signal with low signal-to-noise ratio, the weighted kurtosis The indicators have certain limitations, and parameter optimization is difficult, which affects the detection results

Images