Method Of Transformer Excitation Inrush Current Identification Based On Artificial Neural Network

Electric power is the antecedence to the development of the society.With the rapid development of the national economy,our need for the electric power become increasingly pronounced,which leads the capacity of the power system constantly increasing,and so does our requirement to the quality of power products.Power transformer is very important equipment in the power system that plays a key role in power transmission.Now the transformer of large capacity is widely used,which put forward higher requirement for the security and reliability for the power transformer protection.Differential protection is always the main protection of the transformer,and the actual recognition method for the inrush current can not meet the requirement of modern large transformer now.Thus,the recognition problem has apparently become the core problem of the differential protection of the transformer that urgently needs to look for a new recognition method,which is faster,more accurate,and more reliable.Combined with the technical project of fault analyze system based on wave recording and protection information of the substation in Liaocheng power supply company of Shandong electric power corporation,the research work is conducted,in which the author of the thesis is in charge of the fault diagnosis of the large transformer in the substation.The BP neutral Onetwork is a multilayer feed forward neural network,which can realize the nonlinear mapping between the input and the output,better approximation and self learning ability.The structure is simple and the training speed is fast.Based on the model of BP neural network,the genetic algorithm is applied to optimize the weight and threshold,which can improve the convergence rate of the neural network.The inrush current and the inner fault data gathered is taken as the training sample and the test sample after the spectrum analysis and normalization.The training samples are used to train the model to endow the model the ability of distinguishing the inrush current and inner fault.Then the training effect is proved by the test sample,and the BP neural network and the genetic algorithm model are compared to give judgment in different conditions.The improved BP neural network model based on genetic algorithm is proved to be more quick and accurate in distinguish of inrush current and the inner fault,which is more applicable in the onsite transformer protection.