Fault Diagnosis Of Power Communication Network Based On Association Rules Mining

With the continuous development of power grids,the scale of power communication network and the service carried have been growing.Early fault handling methods have been unable to meet the increasing requirements for network fault management.In addition,the amount of alarm data generated by the network management system every day is huge and complex,and the network nodes and devices are interrelated,So when a network or equipment malfunctions,it will make the related network elements or equipment produce a large number of derivative alarm.A large number of repeating and useless alarm information will submerge root failure,which seriously affects the maintenance personnel to locate the root causes of failure and determine the cause of the malfunction.In order to find the root cause of failure and deal with the fault more quickly and accurately,it is urgent to propose a quick and intelligent method to analyze the alarm information,find out the correlation between them,and then determine the root cause of the fault.Alarm association rule mining is an inductive learning method based on historical alarm database.By mining the alarm information of the database,it can automatically discover the correlation pattern of the alarm data and finds the valuable alarm information.Then we can infer the root cause of the failure.According to the characteristics of alarm data of power communication network,this paper studies the power communication network alarm correlation and fault diagnosis.The concrete research contents and are as follows:In order to convert original alarms to the suitable data for mining,a sliding window method is proposed.This method deletes redundancy data and solves the problem of non-synchronization of alarm time.Meanwhile,a new more scientific and reasonable method based on wavelet neural network is proposed,which can judge the importance of the combination of alarm attributes quickly.Aiming at the traditional association rule mining method has many problems such as insufficient memory and low efficiency when alarm data is large,this paper proposes a parallel mining algorithm for association rules based on MapReduce framework.In addition,the paper discusses the function of the alarm association rules for the maintenance personnel to determine the root failure.Lastly,we analyze the rationality and applicability of the alarm association rules based on the topology structure of the power communication network.