Early Warning Technology Of Lightning Arrester Defects Based On Bayesian Network

With the development of economy,the power system that provides energy for all walks of life is becoming more and more complex,which poses huge challenges to the stability of grid operation and power supply reliability.Among them,lightning arresters are important overvoltage protection equipment in substations.Its defect-free,safe and stable operation is extremely important to the reliability of power supply.Online monitoring of arresters is an important way to evaluate their status Continuous enrichment and improvement of online monitoring systems are also in line with the development direction of modern smart grids for self-healing.After basic online monitoring of arresters,state assessment and defect warnings are carried out accordingly.It is an important means to discover its defects in time.Based on the massive data obtained from actual engineering projects,this article uses Bayesian network and other artificial intelligence methods to develop the technology of analyzing and predicting the state of the arrester and warning of defects,which improves and enriches the online monitoring of the arrester.System,and realized the application in the actual project.First introduce the basic working principle of metal oxide arrester,analyze its defect mechanism and defect factors.On this basis,select the key characteristic quantities that reflect its operating conditions or characterize its defects,and discuss how to obtain these in actual engineering projects.The method of characteristic quantity data takes into account the problems of interference and large errors when measuring leakage current characteristic quantity data in engineering practice.Combined with the patent,a leakage current sensor that eliminates external electromagnetic interference is proposed.The temperature data,the increment of the number of movements and the method of obtaining meteorological data are briefly introduced.Then,information modeling is carried out on the characteristic quantity of the arrester to form a multi-source and multi-dimensional data model.Time series,difference algorithm,and ARMA model are used to predict its key characteristic quantity.Based on this,the defect of the arrester based on Bayesian network is proposed.Early warning algorithm,based on a brief introduction to Bayesian method,discards the advantages and disadvantages of traditional Bayesian method in defect analysis,introduces Bayesian network to early warning its defects,and details the algorithm flow.Finally,discuss the application of the lightning arrester defect early warning algorithm.The Bayesian network-based lightning arrester defect early warning algorithm is applied to the actual project of the Jinan Quancheng Substation lightning arrester online monitoring.The multi-channel source of the arrester data is introduced first,and then the data set is constructed;then the defect is applied The early warning algorithm constructs a Bayesian network model for defect early warning analysis,and uses the model to analyze and early warning actual engineering cases;then,for the shortcomings found in the application test of the model,an optimized Bayesian decision network model is proposed and compared with the preliminary model Comparing the application effects of the application,the automation level of the application is improved,and accuracy of the algorithm model are verified;finally,the sensitivity adjustment of the optimized application model is discussed based on the changes in the power grid security requirements.