Research On Log Analysis And Fault Early Warning In Full-service Unified Data Center Of Electric Power System

State Grid Corporation is carrying out the construction of the full-service unified data center of electric power system,in order to realize the management and analysis of unified storage of all-service scope,all-data type and all-time dimension data.At present,19 sets of business information systems have been connected to the full service unified data center in Sichuan Electric Power Company of China National Grid,and the total amount of data access has reached 110.92 TB.the full-service unified data center of electric power system provides an important technical support and guarantee for the safe and stable operation of the electric power system,and also provides a good guidance for the expansion of the electric network business.With the development of smart grid,the demand for log data acquisition and analysis of various information systems in national grid has increased dramatically.The rapid development of big data technology provides a technical basis for data analysis service of power full-service unified data center.The use of big data technology to obtain potentially useful information from massive real-time log data has been widely concerned.Firstly,this paper expounds the background of the construction of the full-service unified data center of electric power system and its important position in the national power grid information system.Then it introduces the overall structure of the full-service unified data center of electric power system,and specifically analyses the functions of each module of the system.For structured data and measurement unstructured data,corresponding data acquisition schemes and storage methods are given respectively.At the same time,according to the system architecture and log characteristics of the electric power full service unified data center,a log acquisition and processing architecture under the large data environment is designed,and the corresponding log acquisition,storage and log analysis scheme is given.It can effectively collect system logs and analyze data mining by stream processing and batch processing.Build a large data platform environment,and carry out log data acquisition and analysis.Aiming at the fault early warning of the logs of the full-service unified data center of electric power system,an integrated learning prediction algorithm and a fault early warning model based on the improved AdaBoost algorithm are established.Through the experimental analysis of the log data stored by Sichuan Electric Power Company Information and Communication Company,the results show that comparedwith traditional machine learning algorithm,the integrated learning prediction algorithm can always achieve the best prediction effect in different data sets.This paper presents a log data acquisition and processing scheme based on streaming processing,which can efficiently collect and process the system logs of the full-service unified data center of electric power system.Through the fault early warning model based on improved AdaBoost algorithm proposed in this paper,it can effectively carry out system fault early warning through system log analysis,improve the availability of power information system,enhance the security of power information system,and provide necessary support for maintenance related work.