The Application of BP Optimized by a Fuzzy-Controlled Niche Genetic Algorithm on the Prediction of the Fault of Transformer

As genetic algorithm can´t maintain the diversity of groups and is prone to prematurity, in this paper one introduce niche technology and fuzzy control theory to improve the simple genetic algorithm. The niche technology preserve the variety of the population and fuzzy control can make sure GA gets the optimal solutions in the global search by controlling the crossover- probability and mutation-probability. The improved genetic algorithm was used to optimize BP neural network. The optimized BP neural network was used on the prediction of gas-in-oil of a transformer. The results of the experiments show that fuzzy-controlled niche genetic algorithm can avoid the premature convergence effectively during the evolution and improve the accuracy of the prediction. This validates the proposed scheme has some certain practicality and referenced value.