Study On Box Transformer Substation Fault Diagnosis Based On Data Mining Under Iot Architecture

Box transformer substation as an important transformer equipment in distribution system,its reliability affects the safety of the people’s electricity consumption.For a long time,the fault diagnosis of box transformer substation is carried out by manual inspection on site after fails,which is difficult to determine the real cause of the box transformer substation failures.In order to change the traditional fault diagnosis form and realize on-line fault diagnosis,a fault diagnosis model of box transformer is established by acquiring the data under the framework of Internet of Things and mining the mode of box transformer substation fault in the data to establish fault diagnosis model.This on-line fault diagnosis method can eliminate abnormal behavior in the process of box transformer operation in time,prevent system fluctuation and catastrophic accidents.It has important theoretical and practical significance in engineering.This thesis establishes a fault diagnosis framework for box transformer substation based on Internet of Things,analyses the characteristics and working mode of box transformer substation,common faults and their impact on power supply system,builds a three-tier fault diagnosis framework for it,and analyses on-line fault diagnosis process based on Internet of Things,as well as fault mode analysis,evaluation index establishment,fault diagnosis model establishment and self-determination.An improved recursive fault tree and fault mode and impact analysis(FTA-FMEA)method is proposed to study the problem of failure mode analysis and determine the key components that lead to box transformer substation failure.According to the hierarchical structure characteristics of box transformer substation,the recursive FTA-FMEA method is used to clarify the fault logic relationship among different levels of box transformer substation,and the fault evaluation index results of box transformer substation components are obtained.In order to make the evaluation index more objective,the fuzzy comprehensive evaluation method is introduced to improve the evaluation results,and the sequencing results of key fault equipment in the box transformer substation are obtained,and six key fault components are selected as the focus of fault diagnosis and analysis of box transformer substation.A fault diagnosis method based on variable precision rough set and radial basis function neural network(VPRS-RBF neural network)is proposed to establish diagnosis model between key fault components and fault feature parameters in the framework of the Internet of Things.Mine the relationship between fault modes and six key fault devices from 24 feature parameters in order to establish a fault diagnosis model.The variable precision rough set is used as the prenetwork,the radial basis function neural network as the post-network,and the characteristic parameter fault data as the training samples.The algorithm training is carried out to obtain the box transformer substation fault diagnosis model,and the validity of the model is tested by the test samples.A fault type discrimination algorithm based on particle swarm optimization and support vector data(PSO-SVDD)is proposed,which makes the box transformer substation fault diagnosis model have the ability of fault type discrimination and model self-adaptive updating.This method is used to judge whether the fault type belongs to known fault or unknown fault.PSO-SVDD uses particle swarm optimization to optimize the key parameters in support vector data,finds the optimal combination of parameters,trains the optimal SVDD judgment model through the known fault sample data,and uses the unknown fault type test data to judge the validity of the model.The unknown fault type data and the known fault type data are retrained to get the updated fault diagnosis model of box transformer.The validity of the model is verified by the test data.