Citation: Yu, W.; Yu, R.; Tao, J. An
Unsupervised Mutual Information
Feature Selection Method Based on
SVM for Main Transformer Condition
Diagnosis in Nuclear Power Plants.
Sustainability 2022, 14, 2700. https://
doi.org/10.3390/su14052700
Academic Editors: Luis Hernández-
Callejo, Sergio Nesmachnow and
Sara Gallardo Saavedra
Received: 25 January 2022
Accepted: 16 February 2022
Published: 25 February 2022
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
iations.
Copyright: © 2022 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
Article
An Unsupervised Mutual Information Feature Selection
Method Based on SVM for Main Transformer Condition
Diagnosis in Nuclear Power Plants
Wenmin Yu
1
, Ren Yu
1,
* and Jun Tao
2
1
School of Nuclear Science and Techniques, Naval University of Engineering, Wuhan 430034, China;
hust_ywm2012@163.com
2
China Nuclear Power Operation Management Co., Ltd., Jiaxing 314300, China; taoj@cnnp.com.cn
* Correspondence: 18071068480@163.com
Abstract:
Dissolved gas in oil (DGA) is a common means of monitoring the condition of an oil-
immersed transformer. The concentration of dissolved gas and the ratio of different gases are
important indexes to judge the condition of power transformers. Monitoring devices for dissolved
gas in oil are widely installed in main transformers, but there are few recorded fault data of main
transformers. The special operation and maintenance modes of main transformers leads to the fault
modes particularity of main transformers. In order to solve the problem of insufficient samples and
the feature uncertainty, this paper puts forward an unsupervised mutual information method to
select the feature verified by the optimized support vector machine (SVM) model of particle swarm
optimization (PSO) method and tries to find the feature sequence with better performance. The
methos is validated by data from nuclear power transformers.
Keywords:
main transformer; condition monitoring; unsupervised mutual information; feature
selection; DGA
1. Introduction
Power transformers that work under harsh environments would experience thermal
decomposition of oil and cellulose insulation materials, such as arcing, corona discharge,
low energy sparks, severe overloading, overheating of insulation systems and pump motor
failures. These conditions alone or in combination can produce combustible and noncom-
bustible gases [
1
] Detection of anomalies requires an assessment of the amount of gas
produced. Gas in oil-immersed transformers can be used to identify fault types, including
thermal and electrical interference. Gases obtained from chromatographic analysis of insu-
lating oils may contain dissolved carbon monoxide (
CO
), carbon dioxide (
CO
2
), nitrogen
(
N
2
), hydrogen (
H
2
), methane (
CH
4
), acetylene (
C
2
H
2
), ethylene (
C
2
H
4
), and ethane (
C
2
H
6
).
The composition, formation rate and specific content ratio of dissolved gas can be used to
indicate transformer condition.
The composition and content of dissolved gases in oil of transformer insulation can
reflect the operation condition of transformer to a great extent thus dissolved gas analysis
(DGA) has become an effective method for fault diagnosis of oil-immersed transformers [
2
].
Organizations such as the Institute of Electrical and Electronics Engineers (IEEE) and
the International Electrotechnical Commission (IEC) recommend a variety of diagnostic
techniques [
3
], depending on the type of transformer and operating conditions. Some of the
most commonly used techniques include Doernenburg ratio, Rogers ratio, Duval triangle
model, etc. These classical diagnostic methods mostly take the ratio of different gases as the
characteristic input and then judge the actual operating condition of the transformer by the
threshold value formed by experience or statistical methods. Fuzzy network, support vector
machine, artificial neural network, and other commonly used artificial intelligence methods
Sustainability 2022, 14, 2700. https://doi.org/10.3390/su14052700 https://www.mdpi.com/journal/sustainability
