Faulty Diagnosis System of Traction Transformer Based on Genetic Algorithm and Fuzzy Neural Network

Li Na,Liu Mingguang,Yang Gang
DOI: https://doi.org/10.13382/j.jemi.2008.05.001
2008-01-01
JOURNAL OF ELECTRONIC MEASUREMENT AND INSTRUMENT
Abstract:Aiming at the common faults of traction transformer, a new faulty diagnosis method is proposed in this paper, which combines Genetic Algorithm (GA) and Fuzzy Neural Network (FNN). Considering the characteristic that the fault signal boundary is absolutely clear, fuzzy mathematics is used to pre-process the fault signal. To overcome the deficiency of traditional BP algorithm such as slow convergence rate and easily converging at a local minimum, Genetic Algorithm is applied to optimize the initial value of FNN structure. Simulation result shows that this new algorithm can improve the convergence rate and reduce the training time, and the trained GA-FNN is effective in faulty diagnosis of power transformer and the faulty diagnosis accuracy can reach as high as 99%.
What problem does this paper attempt to address?