Title: Detection method of abnormal state of power transformer based on SOM neural network

Authors: Wei Tong; Qi-Ping Huang

Addresses: Anhui Wenda Information Engineering College, School of Computer Engineering, Anhui, Hefei, 231201, China ' Electrical Engineering Department, Anhui Electrical Engineering Professional Technique College, Anhui, Hefei, 230051, China

Abstract: In order to reduce the detection error of abnormal states of power transformers, a power transformer abnormal state detection method based on SOM neural network is proposed. Firstly, collect data such as current, voltage, temperature, and vibration of power transformers, and extract the characteristics of abnormal states of power transformers through empirical wavelet transform. Secondly, by constructing a decision function, the pre-processing results of abnormal state detection feature quantities for power transformers are obtained. Finally, based on the pre-processed abnormal state features and the detection results of power transformer abnormal states, a power transformer abnormal state detection model is constructed using SOM convolutional neural network. The experimental results show that the proposed method is relatively superior in terms of anomaly detection error, with a detection error of only 4.78 cm under a 5% delay error.

Keywords: SOM neural network; power transformer; abnormal state detection; empirical wavelet transform; EMD.

DOI: 10.1504/IJMIC.2024.142274

International Journal of Modelling, Identification and Control, 2024 Vol.45 No.2/3, pp.154 - 163

Received: 27 Oct 2023
Accepted: 29 Feb 2024
Published online: 16 Oct 2024 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article