Prediction Model for Breakdown Voltage of Transformer Oil Based on Relative Transformation and Kernel Principal Component Analysis

Authors

Yinguo Xiong, School of Physical Science and Engineering, Yichun University, Yichun 336000, China;

Abstract

Abstract: Aiming at the difficulty of measuring the breakdown voltage of transformer oil on line, a new prediction model for breakdown voltage of transformer oil is proposed based on relative transformation (RT) and kernel principal component analysis (KPCA). By analyzing the factors that are closely related to the breakdown voltage, the original data space is converted to the relative data space by relative transformation to improve the distinguishability between data. KPCA is employed in the relative space for the purpose of data dimension reduction, denoising and extracting nonlinear features. Kernel principal components extracted by KPCA are used as the input of kernel extreme learning machine (KELM) to establish the prediction model for breakdown voltage of transformer oil, and the parameters of prediction model are optimized by differential evolution algorithm. Compared with RTKPCA-LSSVM, RTPCA-KELM and RTPCA-LSSVM, the simulation results illustrate that the proposed prediction method has better prediction precision and generalization ability.

Recommended Citation

Xiong, Yinguo (2019) "Prediction Model for Breakdown Voltage of Transformer Oil Based on Relative Transformation and Kernel Principal Component Analysis," Journal of System Simulation: Vol. 30: Iss. 5, Article 5.
DOI: 10.16182/j.issn1004731x.joss.201805005
Available at: https://dc-china-simulation.researchcommons.org/journal/vol30/iss5/5

First Page

1657

Revised Date

2016-08-12

DOI Link

https://doi.org/10.16182/j.issn1004731x.joss.201805005

Last Page

1664

CLC

TP183

Recommended Citation

Xiong Yinguo. Prediction Model for Breakdown Voltage of Transformer Oil Based on Relative Transformation and Kernel Principal Component Analysis[J]. Journal of System Simulation, 2018, 30(5): 1657-1664.

DOI

10.16182/j.issn1004731x.joss.201805005