A Fault Diagnosis Method for Food on-load Tap Changer Based on Probabilistic Neural Network

摘要

In this study, we present a fault diagnosis method based on Probabilistic Neural Network (PNN) to find the food on-Load Tap Changer (FFOLTC)s' faults. First the sample data was collected from the results of AC dynamic characteristic tests of (FFOLTC)s. Second features was extracted from the sample data and normalized. Then the parameters were set for the PNN and the samples were trained to get the diagnosis network. Finally we used the test data of FFOLTC to check the network for diagnosis. Experimental results show that the PNN method could detect the complex relationships, could be developed basis for the FFOLTC test data that can identify the fault types. The accuracy of the results is more than 70% in all cases and 100% in some cases. So the proposed method is fast, accurate, easy to modify and can be easily applied to practical application.