摘要
在变压器故障诊断中,针对样本中可能存在的孤立样本以及样本分布的不均匀性,导致神经网络整体性能下降,训练和测试效率降低的情况,设计了利用模糊聚类法对样本进行预先处理,然后再应用神经网络进行训练和测试的诊断系统.从仿真结果可以看出,该诊断系统较样本未处理而直接应用于神经网络诊断系统的诊断性能有大幅度的提高.
In fault diagnosis system of power transformer, when there are some isolation samples in the sample set or when the distribution of sample is not symmetrical, it may cause performance of the diagnosis system and lower the efficiency of training and testing. fuzzy clustering method is introduced to overcome these problems by preprocessing damage to the In this paper, the sample set before it is used to train the diagnosis system. The results of simulation prove that the proposed method is effective in solving the problems and enhancing the performance of fault diagnosis system.
出处
《上海电力学院学报》
CAS
2010年第2期182-185,189,共5页
Journal of Shanghai University of Electric Power
关键词
模糊聚类
神经网络
变压器
故障诊断
fuzzy clustering
neural networks
transformer
fault diagnosis
