978-1-5386-6057-7/18/$31.00 ©2018 IEEE
Cuckoo Search Optimized NN-based Fault Diagnosis
Approach for Power Transformer PHM
Anyi Li
College of Information Engineering
Nanchang University
Nanchang, China
lianyi@ncu.edu.cn
Huanyu Dong
Nanchang University
Nanchang, China
donghuanyu@ncu.edu.cn
Xiaohui Yang*
College of Information Engineering
Nanchang University
Nanchang, China
*Corresponding author: yangxiaohui@ncu.edu.cn
Chunsheng Yang
National Research Council Canada
Ottawa, Ontario Canada
Chunsheng.Yang@nrc-cnrc.gc.ca
Abstract—An emerging prognostic and health management
(PHM) technology has recently attracted a great deal of attention
from academies, industries, and governments. The need for
higher equipment availability and lower maintenance cost is
driving the development and integration of prognostic and health
management systems. PHM systems enable a pro-active fault
prevention strategy through continuously monitoring the health
of complex systems. Power transformer PHM will play a key role
in securing and stabling electrical power supply to users,
especially in the smart grid. In this paper, we present a novel
approach for power transformer fault diagnosis based on cuckoo
search optimized neural network, also named it as dissolved gas
analysis (DGA) approach. The proposed approach uses the
Cuckoo Search (CS) algorithm to select the best parameters of
backpropagation (BP) neural network, which can approximate
any nonlinear relationships. The paper validates the usefulness
and efficiency of the proposed approach by conducting
simulation to compare the results to Particle Swarm
Optimization (PSO) and Genetic algorithm (GA). The results
demonstrated that the proposed approach outperformed other
methods such as BP neural network, SVM, GA-BP, and PSO-BP.
It significantly improved the performance and accuracy of fault
diagnosis/detection for power transformer PHM.
Keywords— Power Transformer, Cuckoo Search, BP Neural
Network, multi-hidden layer, Fault Diagnosis
I. INTRODUCTION
An emerging prognostic and health management (PHM)
technology has recently attracted a great deal of attention from
academies, industries, and governments. The need for higher
equipment availability and lower maintenance cost is driving
the development and integration of prognostic and health
management systems. Taking advantage of advances in sensor
technologies, PHM systems enable a pro-active fault
prevention strategy through continuously monitoring the health
of complex systems. Power transformer PHM will play a key
role in securing and stabling electrical power supply to users,
especially in the smart grid. Transformer is the most important
electrical equipment in electrical power system [1, 2]. The
operating state of transformer is directly related to the stability
of whole system [3]. Therefore, it is of vital importance to
monitor the condition of transformer and investigate the
potential fault in order to improve the accuracy of fault
detection.
Dissolved gas analysis (DGA) [4, 5] has been widely
recognized as an efficient method in power transformer faults
detection. Based on color phase chromatography this method
can analyze the gas content dissolved in the oil. It can
determine whether a transformer is abnormal by the
composition of gas in the oil and the content of various gases.
However, power transformer structure is very complex, there
are many characteristic quantities to represent its status, and
there is huge uncertainty and fuzziness between the status
messages, it creates much difficulty to the fault diagnosis [6].
The methods of power transformer faults diagnosis [7]
include mainly IEC four-ratio and the three-ratio method,
characteristic gas method and so on. But these methods
generate large errors in the diagnosis of power transformers.
The accuracy will be greatly reduced when the sample data is
too large or there is some interference among the samples.
Therefore, the artificial intelligence technology with excellent
performance is desired to be used in transformer fault
diagnosis. The intelligent algorithms based on the DGA data
are the widely-used methods in transformer fault diagnose,
especially the backpropagation (BP) neural network [8, 9]. The
BP neural network can be used to find out the connection
weights and bias to implement accurate diagnostic methods or
models for DGA. The updated parameters of BP neural
network follow the rule of gradient descending to avoid
mistaking the parameters as the optimal parameters.
This work was supported in part by the National Science Foundation o
China (51765042, 61463031, 61662044, 61773051), Jiangxi Provincia
Department of Science and Technology JXYJG-2017-02.