On a new improved prediction algorithm employed in the fault transformer

Although simple genetic algorithm (SGA) can, to some extent, improve the back propagation neural network (BP), it is prone to prematurity and losing the optimal solutions. Niche technology and fuzzy control theory are introduced to improve SGA and the improved one is used to optimize BP. The improved genetic algorithm is used to optimize BP neural network. In addition, due to the increasingly voltage levels and the effect from many other uncertain factors such as the continuously changing temperature, the application of a single forecasting model is limited. So in the end of this paper, the BP optimized is combined with GM algorithm, which was proposed by the known professor Julong Deng in 1982 and is popular with the researchers studying prediction. Both of the optimized BP and the combinational predicting model was used on the prediction of gas-in-oil in some transformers. The results of the experiments show that the proposed optimizing strategy is valuable and practicable.