Transfomer fault diagnosis based on rough sets theory and artificial neural networks
Xiaodong Yu, Hongzhi Zang · 2008
Transformer fault diagnosis based on artificial neural networks (ANN) is widely used, because ANN has essential nonlinear character, parallel processing ability and the ability of self organize and self learning. But there exist problems if we use traditional ANN method alone to diagnose transformer fault, the large input vector dimension and complex training database will cause the computation complexity and the space increase greatly, lead to long training time, slow convergence and low judgement accuracy. In this paper, a hybrid fault diagnosis method combining rough set (RS) theory and ANN (RS-ANN) is presented. Taking advantage of the strong ability of RS theory in processing large data and eliminating redundant information, this method can remove irrelevant factors from the original data and reduce the amount of training data which helps to overcome ANN's defect when process large database. A number of simulation results show that RS-ANN simplifies the networks' structure, reduces the networks' training epochs, improves the judgement accuracy.