Fault Diagnosis Of Power Communication Network Based On Alarm Correlation

The power communication network management system is an important guarantee to ensure the safe and stable operation of the power communication network.When the physical resources in the network management system fail,the network management system will receive the relevant alarm information indirectly at the first time and arrange the staff to deal with it.This paper focuses on the problems related to the alarm information of power communication network management system,and divides the work into three aspects: data preprocessing of network management system,fault diagnosis of network management resources and alarm information level of network management resources.1.The alarm data preprocessing based on power communication network management system is studied.Because the expression form of the alarm information in the network management system is not uniform and the root alarm cannot be judged quickly,the root alarm in the original alarm data of the network management system can be obtained by using the alarm correlation analysis method from the occurrence of the alarm to the elimination of the alarm.For the missing alarm information caused by transmission failure in the process of data transmission,the root alarm data samples with missing items are classified by decision tree,and the missing values are filled by decision tree classification to ensure the integrity of the data,which is verified by experiments.2.The fault diagnosis scheme of power communication network based on network management system deep belief network is designed.Considering that deep belief network uses gradient descent method to update the parameters in fault diagnosis of network management resources and achieve the purpose of learning.In order to solve the problem that the sample data cannot be effectively learned due to gradient descent,the activation function of each hidden layer selected in the deep belief network is changed to ensure the diversity of data and avoid the problem that the parameters are not updated due to gradient descent.Inthe experiments,the alarm information network management system of six types of physical resources as the basis of fault diagnosis,the relevant properties(such as the type of alarm level,alarm and alarm causes)as input data,the characteristics of the data can be divided into training set and test set,training and learning through the training set and test set authentication model of fault diagnosis performance.3.The alarm level division scheme based on Bayesian network of network management system is designed.In the network management system,when there are multiple alarm messages,the alarm level caused by the fault is prioritized and sorted,and the fault is processed in order.Considering the inclusion relation of physical resources,the fault tree is constructed by analyzing the alarm information,the logical relation between the physical resources is analyzed,and the Bayesian network is generated to realize the level division of the warning information of power communication network,and its effectiveness is verified by comparative experiments.