The Method Of Improved BP Algorithm For Genetic Simulated Annealing Algorithm In Grounding Fault Line Selection

In the medium and low voltage distribution network of our country,single-phase ground faults are the most common.Although the distribution network can maintain normal operation within a short period of time after a ground fault occurs,long-term operation would cause serious problems such as multi-phase short circuits.It is necessary to troubleshoot and repair the faulty line as soon as possible.Therefore,ground fault detection has important significance and research value for the safe operation of distribution networks.Due to the complex structure of the distribution network circuit,the fault signal characteristics are not obvious,and it is easily affected by many factors such as grounding resistance,grounding phase and grounding mode.In the traditional fault line selection method,the accuracy of the single criterion of the fault feature is not high,and it is very easy to be affected by different factors to cause misjudgment;while the conventional mathematical formula is used to determine the fit of multiple criteria.It is complex and lacks versatility for failures of different structures and conditions.With the development of artificial neural networks,artificial intelligence can be used for multi-criterion fusion to select ground fault lines.Among them,the BP neural network can effectively perform data training and discrimination,but its training efficiency is low,and it is extremely susceptible to the initial weight and threshold of the neural network.Aiming at the shortcomings of BP neural network,genetic simulated annealing algorithm is used to optimize its data,improve convergence efficiency,enhance global search ability,and eliminate the influence of excessive initial weight threshold.To the problem that genetic algorithm is easy to fall into local minimum and the convergence speed of simulated annealing algorithm is slow,an improved genetic simulated annealing algorithm is used to optimize the line selection method of BP neural network(IGSAA-BP).This algorithm not only avoids the problem that the traditional BP neural network has too high proportion of the initial weight threshold,but also improves the diversity of the population by improving the cross-mutation probability formula of the genetic algorithm and the Metropolis criterion to avoid falling into the local optimum.Build a fault simulation model to obtain the active characteristics,5th harmonic characteristics and transient characteristics of the zero-sequence current.Input the IGSAA algorithm to obtain the initial value of the BP neural network,and then obtain the line selection result after the neural network model training calculation.Compared with other BP neural network algorithms,it shows that this method has higher convergence speed,higher complexity adaptability and better judgment accuracy in training.