The Research Of SDH Alarm Association Rules And Fault Diagnosis In Power System Communication Network

The role of elctric power communication network in the electric power system isvery important, especially in the security and stability control systems of electric powerand dispatching automation systems, commercialization of the electricity market operator,achieving manage the electric power system modernly and diversification in thenon-power industry.It plays a base, indemnification and premise role. In order to ensurethe stability of the power of communication to run, When fault occurs, people need tomake the right judgment of specific issues of the failure in the shortest possible time, andclear the trouble immediately. With increasing degree of complexity of the network, Theefficiency of the original manual analysis is very low. As the electric powercommunication network has a complex interconnection in physic and logic, a singledevice alarms or failure is often resulted by the associated equipment affect each other.Analysis these alarm’s correlation to find out the implicit association rules, use theserules can shield unnecessary or irrelevant alarms, reduce the amount of alarms reportsand improve the efficiency of fault diagnosis.This paper describes the SDH (Synchronous Digital Hierarchy) concepts and SDHnetwork management system firstly, and did a deep discussion and research incharacteristics of SDH alarms and the alarm process. After deletes the redundant andmerge the important fields, use the method of dynamic sliding window on the alarm datato devide the data in order to form the alarm transaction set, and use the SQL language toachieve the same alarm transaction repeat alarm within the group only to return one.After the alarm after data preprocess, the use of data mining components in the MicrosoftSQL server2008to build the mining model, called Microsoft association rules algorithm,according to the selected unified format the data in the table of data mining, appropriateadjustment of mining parameters, output mining results, after analyzing reasonably cometo the association rules between between alarm and alarm types.Paper also implements the Apriori algorithm for mining association rules in thesimulation experiment, for mining frequent itemsets at the time of scanning database formany times, to generate the data size is large, the disadvantage of a large number oftime-consuming and memory overhead, by reducing the way of scan data scale, using the method of multistage support gives the LIG(Large Item Generation)algorithm, andcompared the two algorithms of mining results and mining time.