Scholarlink

The insulation resistance of DC voltage ratio standard devices changes non-linearly when the voltage is increased, causing uneven potential distribution in the measured resistance layer, which in turn leads to a reduction in the measurement accuracy of the standard devices. The mechanism of nonlinear change of insulation resistance is complicated, this paper fully considers the influence of temperature and humidity, voltage level, Granding ring parameters and other factors. The first step is to build an experimental platform to obtain the data set of the effects of these factors on the nonlinear characteristics of the insulation resistance, and then construct a mathematical model of the nonlinear characteristics of the insulation resistance. Then designing an AFBP neural network to established an insulation resistance nonlinear prediction model. Finally, the model is trained and fitted based on 2558 sets of data on nonlinear characteristics of insulation resistance under different influencing factors and compared with other machine learning algorithms. The experimental results show: the AFBP-based insulation resistance prediction model has optimized the MSE and other indexes by 10%– 50% compared with other machine algorithms, and is able to estimate the magnitude of insulation resistance under different influencing factors in a better way.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Research on the Prediction of Nonlinear Characteristics of Electrical Insulation Resistance of Standard Devices Based on AFBP Neural Network

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Yin Gao,
  • Chao Wang,
  • Ran Gao,
  • Zheng Chen,
  • Jie Shi,
  • Minlong Li,
  • Gujing Han

摘要

The insulation resistance of DC voltage ratio standard devices changes non-linearly when the voltage is increased, causing uneven potential distribution in the measured resistance layer, which in turn leads to a reduction in the measurement accuracy of the standard devices. The mechanism of nonlinear change of insulation resistance is complicated, this paper fully considers the influence of temperature and humidity, voltage level, Granding ring parameters and other factors. The first step is to build an experimental platform to obtain the data set of the effects of these factors on the nonlinear characteristics of the insulation resistance, and then construct a mathematical model of the nonlinear characteristics of the insulation resistance. Then designing an AFBP neural network to established an insulation resistance nonlinear prediction model. Finally, the model is trained and fitted based on 2558 sets of data on nonlinear characteristics of insulation resistance under different influencing factors and compared with other machine learning algorithms. The experimental results show: the AFBP-based insulation resistance prediction model has optimized the MSE and other indexes by 10%– 50% compared with other machine algorithms, and is able to estimate the magnitude of insulation resistance under different influencing factors in a better way.