A multi-dimensional information fusion method for transformer fault diagnosis based on depth learning

Abstract

The invention provides a multi-dimensional information fusion transformer fault diagnosis method based on depth learning. Firstly, the multi-dimensional state monitor which can best reflect the real operation state of the transformer is selected as the input of the depth learning model for normalization. Secondly, the common transformer faults that need to be identified are coded and classified. Then, based on the depth learning theory, sparse constrained Boltzmann machines are stacked to form the underlying network. Finally, a classifier is added to the top of the depth learning model to forma depth learning model based on sparse depth belief network. The fault diagnosis method can make use of a large number of unlabeled multi-dimensional monitoring data of transformer as learning samples, and only a small number of labeled multi-dimensional monitoring data can be used for optimization. The trained model can make accurate diagnosis of transformer status according to the real-time multi-dimensional monitoring data of transformer. The fault diagnosis method provided by the invention is feasible and effective, and the performance thereof is superior to the existing transformer faultdiagnosis method.

Description

technical field [0001] The present invention relates to research on transformer fault type diagnosis, and proposes a deep learning-based multi-dimensional information fusion transformer fault diagnosis method. Background technique [0002] As the most core power equipment in the power transmission and transformation system, the operation status of the transformer is directly related to the safe and stable operation of the entire power grid. Once a failure occurs, it will inevitably cause local or even large-scale power outages, resulting in huge economic losses. There are many reasons for the failure of transformers. Therefore, it is of great significance to improve the safe and stable operation of transformers by quickly and accurately judging transformer failures through a large number of multi-source heterogeneous monitoring data. [0003] At present, the fault diagnosis methods of power transformers are mainly divided into traditional methods and intelligent methods. T...

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to propose an intelligent transformer fault diagnosis method. By establishing a semi-supervised learning algorithm, a large number of unlabeled samples are effectively used to pre-train the diagnostic network, which solves the problem that intelligent algorithms such as SVM and neural network are too dependent on labels. Sample problems, with stronger learning ability and better diagnostic performance

Images