Feature selection using RapidMiner and classification through probabilistic neural network for fault diagnostics of power transformer

The diagnosis of incipient fault is important for power transformer condition monitoring. The incipient faults are monitored by conventional and artificial intelligence based models. The key gases, percentage value of gases and ratio of Doernenburg, Roger, IEC methods are input variables to artificial intelligence (AI) models which affects the accuracy of incipient fault diagnosis so selection of most influencing relevant input variable is an important research area. With this main objective, RapidMiner software is applied to IEC TC 10 and related datasets having different operating life to find most influencing input variables for incipient fault diagnosis in AI models. The RapidMiner identifies %CH₄, %C₂H₂, %H₂, %C₂H₆, C₂H₄/C₂H₆, C₂H₂/CH₄, C₂H₂/H₂ and CH₄/H₂ as the most relevant input variables in incipient fault diagnosis and it is used for fault diagnosis using different artificial intelligence (AI) approach i.e. fuzzy-logic (FL) and . The compared results shows that AI models give better results at proposed input variables used as an input vector. PNN gives highest accuracy of 98.28, proving proposed input variables can be used in transformer fault diagnosis research.