An Intelligent Diagnosis Model of Power Grid Transformer Fault Based on BP Neural Networks and Genetic Algorithms

Grid transformer is the most important facility of power system transmission and transformation equipment, and also one of the electrical equipment with high accident rate. Detecting latent faults of transformers at any time and eliminating hidden problems in time is an effective way to ensure their safe operation and upgrade power supply reliability. To this end, the study introduces genetic algorithms (GA) to the traditional BP network hitch diagnosis method and applies it to the hitch diacrisis of grid transformers. The simulation results show that the overall test data error of Back Propagation (BP) network algorithm is reduced to about 0.1 after adding genetic algorithm. The training time of the GA-BP algorithm is 84.4s, which is significantly shorter than that of the conventional algorithm. The hitch diacrisis precision of the studied model is 95.04
^%
, which is 12.39% and 7.11 % higher than the GA and BP algorithm, respectively. The value of
$F_{1}$
is 96.39% and the error of its hitch diacrisis result is only 0.87%. The research method improves the fault speed and correct diacrisis rate, and better improves the safety of grid transformer operation, which has important practical significance.