摘要
设计了４ 种发电机定子线棒工业仿真模型，取得了不同放电模式（ 类型和发展程度） 的大批试验数据。用φｑｎ三维谱图方式压缩放电信息，ｑ 轴采用对数刻度。以谱图表列数据或曲面拟合参数为放电样本的特征量。为了能更有效地识别放电模式，研究了基于任务分解网络模块的人工神经元网络组。使用前馈网络构成类型识别主网络，网络有足够满意的类型识别率。对程度识别子网络。
Four kinds of industrial simulation models of generator stator winding were designed and a lot of experimental data of different discharge patterns (types and serious levels) were obtained. The discharge information was suppressed through φ q n 3 dimensional pattern and the logarithms scale was used for q axis. The tabulated data or surface fitting parameters of these patterns were used as characteristic vectors. For purpose of more efficiently discriminating discharge pattern the artificial neural network group based on task dissolved net modules was studied. The main network composed of feed forward network had enough recognition rates of type discrimination. For subnetworks used for serious levels recognition the self organization feature map network could more imaginably discriminate the serious levels of discharges.
出处
《中国电机工程学报》
EI
CSCD
北大核心
1999年第10期64-67,71,共5页
Proceedings of the CSEE
关键词
发电机
定子
仿真模型
线棒
放电
模式识别
winding of electrical machine
insulation diagnosis
artificial neural network
pattern recognition
