Power Grid Fault Diagnosis Based On GAN

With the increasingly complex structure and fault characteristics of power grid,fast and accurate fault diagnosis of power grid has become more and more important for the safe and stable operation of power grid.The development and application of artificial intelligence technology provide new ideas and methods for the fault diagnosis of power grid.Most of the current fault diagnosis methods have achieved good results for simple fault diagnosis,but in the face of the lack of developmental fault data and complex fault characteristics,the effective and accurate diagnosis cannot meet the requirements of the actual complex operation of large power grid fault diagnosis.To solve the problem of lack of developmental fault sample data,this paper proposes a method of generating developmental fault samples based on CycleGAN.Generating samples containing noise can better test the diagnostic ability of fault diagnosis model.First,in the face of simple faults,the electrical volume that can reflect the fault characteristics is selected,and the corresponding PMU data is converted into a 4-dimensional polar plot.Then,the simple fault PMU polar plot was used to train and test the developmental fault sample generation model,and the corresponding simple fault to developmental fault samples were generated.Finally,a complete developmental fault sample generation process and diagnostic application strategy are proposed.In view of the difficulty in diagnosing complex faults,a fault diagnosis method based on the fusion of GAN alarm information and PMU data is proposed in this paper according to the complementary characteristics of switching quantity and gas data.This method still adopts the idea of PMU data graphics to convert the fault-related electrical gas data into a 7-dimensional polar plot,and then inputs the alarm information text and PMU polar plot under the same fault into the designed GAN model for data fusion training and output the diagnosis results.The experimental results show that the fusion of dual data source features not only further improves the diagnosis accuracy of simple faults,but also effectively identifies faults in the case of complex faults such as circuit breaker rejection and protection rejection,with a high diagnostic accuracy.