A Noise Separation Method Based on Unsupervised Mode for Working Conditions of Power Transformers

Abstract

The invention relates to an unsupervised method for separating working condition noise of a power transformer, which solves the defect that it is difficult to separate the noise from the voiceprint signal of the power transformer compared with the prior art. The invention comprises the following steps: collection of voiceprint data of power transformers; preprocessing of voiceprint data; compression of transformer voiceprints; separation of transformer working condition noise. In the present invention, through the collected transformer audio, after denoising processing, the frequency compression method and frequency features are used to extract the transformer voiceprint features and effectively fuse them; for the sound of the transformer working condition, the sound emitted by the stable working condition and the unstable transient noise For the additive superposition problem, an algorithm based on the angle similarity between eigenvectors is proposed, which realizes the separation and detection of voiceprint superposition, separates the noise characteristics of transformer working conditions, and lays the foundation for the judgment and detection of transformer working conditions.

Description

technical field

[0001] The invention relates to the technical field of power equipment, in particular to an unsupervised mode-based noise separation method for working conditions of power transformers. Background technique

[0002] Due to the influence of latent fault factors such as internal discharge, overheating, winding deformation, loose mechanical parts and equipment insulation aging of large power transformers (reactors), serious faults will occur during the operation of power transformers (reactors) over time. The traditional manual judgment method is to judge the type and location of the fault by listening to the operating sound of the equipment with the human ear. This type of method has certain limitations, ambiguity, subjectivity and lack of persistence. The specific performance is that different technicians may have inconsistent judgment standards for transformer faults, and some subtle acoustic signals cannot be judged, which requires the accumulation of certai...

Images