Case-based reasoning adaptive optimization algorithm for power transformer fault diagnosis

Wei Zhang, Qiu-li Wu, Yu-rong Deng, Ze-cheng Lv 

Guangxi Electric Power Research Institute, 530023 Nanning Guangxi Zhuang Autonomous Region, China 

The adaptive learning rate for the introduction of case-based reasoning transformer fault type identification. The adaptive learning rate theory, through improved data normalization, typicality and best filtering diversity to extract the original example and optimal neural network. In the sample processing and analysis process to be solved according to the type of fault feature automatically adjusts the data processing methods, processes, boundary conditions and constraints to adapt statistical distribution, the probability characteristics. Examples show that this method can overcome the DGA data ambiguity and dispersion problems in the recognition accuracy and convergence speed advantage.