Research On Data Mining And Application Of Distribution Networks Based On Big Data Platform

With the construction and development of China's intelligent distribution network,the degree of intelligence and informatization of distribution network is constantly improving,and various informatization systems are widely used in distribution network.A large number of operation data of distribution network are recorded in each system,which contains hidden rules of operation and maintenance of distribution network.If the hidden rules can be fully excavated,it can provide support for lean operation and maintenance of distribution network.At present,the data scale of distribution network information system is still increasing sharply,and the data complexity is also increasing.It is difficult to dig out the hidden rules in the process of operation and maintenance by using traditional data mining methods.Large data technology has developed vigorously in recent years.It has shown strong advantages in mining multi-source heterogeneous data,and has been widely used in economic,financial,transportation and other fields.The application of big data technology to distribution network data can efficiently and deeply excavate useful value and assist grid personnel in operation and maintenance decision-making.In this paper,large data technology is applied to data mining of distribution network operation and maintenance to achieve data acquisition,data preprocessing,data mining and visual display of distribution network.This paper focuses on data mining and visualization display from two aspects of distribution network operation and maintenance level evaluation and operation and maintenance event sequence association mining.The main work of this paper is as follows:(1)This paper introduces the key technologies of big data from three aspects:big data platform,big data mining technology and big data visualization technology.The big data platform mainly introduces Hadoop ecosystem,HDFS and MapReduce.Big data mining technology introduces the concept and common classification of big data mining,and focuses on the clustering algorithm and sequence mining algorithm used in this paper.The common large data visualization tools are sorted out and their characteristics are analyzed.(2)The data sources of distribution network are fully investigated and analyzed,and operation and maintenance data are collected from multiple information systems.Based on Hadoop large data cluster environment,according to the specific needs of data mining,data cleaning and fusion for operation and maintenance are carried out.(3)A comprehensive evaluation method of operation and maintenance level based on clustering algorithm and radar image is proposed.Firstly,the shortcomings of K-means algorithm are analyzed,and the initial point selection of K-means algorithm is improved based on the idea of data density.Considering the shortcomings of clustering analysis in visualization and evaluation,the clustering algorithm and radar image evaluation method are combined.Then,aiming at the shortcomings of traditional radar map and the characteristics of operation and maintenance data,the method of drawing radar map and the method of calculating characteristic quantity are improved.Finally,the comprehensive evaluation method is applied to the actual operation and maintenance data to realize the generality mining and quantitative evaluation of operation and maintenance level.(4)Complete operation and maintenance event sequence association mining and visual display.Based on the analysis of the characteristics of operation and maintenance data,a complete sequence partition method is proposed.PrefixSpan is used to sequence,and then parallel coordinate visualization technology is used to solve the problem of visualization of sequence mining results.Finally,the method is applied to the actual operation and maintenance data to mine the internal relationship between the operation and maintenance events.