基于卷积神经网络的变压器故障诊断方法
发布时间：2019-01-24 21:09
【摘要】：变压器是电力系统中的重要设备,其安全与稳定直接影响着国民经济的健康发展。油中溶解气体分析(Dissolved Gas Analysis,DGA)是分析变压器故障类别的重要手段。卷积神经网络是深度学习的一种模型,广泛应用于图像识别、语音处理等领域,具有非常好的分类能力。文章选取了变压器的五种油中溶解气体含量作为模型输入量,在借鉴传统浅层BP神经网络油中气体分析方法的基础上,针对BP神经网络表达能力不足以及容易过拟合的缺点,将卷积神经网络应用于变压器故障诊断,并与BP神经网络的分类效果进行了对比,通过算例研究证明了卷积神经网络的效果更优。文章也对卷积神经网络的卷积核个数、卷积核大小以及采样宽度对分类效果的影响进行了探讨。
[Abstract]:Transformer is an important equipment in power system. Its safety and stability directly affect the healthy development of national economy. Dissolved gas analysis (Dissolved Gas Analysis,DGA) in oil is an important method to analyze transformer fault types. Convolutional neural network is a kind of model of deep learning, which is widely used in image recognition, speech processing and other fields, and has a very good classification ability. In this paper, the dissolved gas content in five kinds of oil of transformer is selected as the input quantity of the model. On the basis of drawing lessons from the traditional gas analysis method of shallow BP neural network oil, the expression ability of BP neural network is insufficient and it is easy to over-fit. The convolutional neural network is applied to transformer fault diagnosis, and compared with the classification effect of BP neural network. The numerical example shows that the convolution neural network is more effective. The effects of the number of convolution kernels, the size of convolution kernels and the sampling width on the classification effect of convolution neural networks are also discussed in this paper.
【作者单位】： 华南理工大学电力学院;
【基金】：国家重点基础研究发展计划(973计划)(2013CB228205) 国家自然科学基金资助项目(51477055)
【分类号】：TM407;TP183
[Abstract]:Transformer is an important equipment in power system. Its safety and stability directly affect the healthy development of national economy. Dissolved gas analysis (Dissolved Gas Analysis,DGA) in oil is an important method to analyze transformer fault types. Convolutional neural network is a kind of model of deep learning, which is widely used in image recognition, speech processing and other fields, and has a very good classification ability. In this paper, the dissolved gas content in five kinds of oil of transformer is selected as the input quantity of the model. On the basis of drawing lessons from the traditional gas analysis method of shallow BP neural network oil, the expression ability of BP neural network is insufficient and it is easy to over-fit. The convolutional neural network is applied to transformer fault diagnosis, and compared with the classification effect of BP neural network. The numerical example shows that the convolution neural network is more effective. The effects of the number of convolution kernels, the size of convolution kernels and the sampling width on the classification effect of convolution neural networks are also discussed in this paper.
【作者单位】： 华南理工大学电力学院;
【基金】：国家重点基础研究发展计划(973计划)(2013CB228205) 国家自然科学基金资助项目(51477055)
【分类号】：TM407;TP183
【相似文献】
相关期刊论文 前10条
1 王雪峰,邬建华,冯英浚,王建元;运用样本更新的实时神经网络进行短期电力负荷预测[J];系统工程理论与实践;2003年04期
2 文汉云;;硫化氢燃烧的神经网络PID控制及其仿真[J];自动化与仪器仪表;2006年01期
3 张U，
本文编号：2414862
本文链接：https://www.wllwen.com/kejilunwen/zidonghuakongzhilunwen/2414862.html
