Transformer Fault Diagnosis And Analysis In Parallel Based On The Spark Cloud Platform

With the rapid development of smart grid,the power industry has entered the "era of big data." Transformer is the key equipment to the grid smooth running,and transformer fault diagnosis methods can guarantee the power system running smoothly.In the power system,the use of transformer on-line monitoring technology can be found that the fault type in a timely manner.But as the monitoring points are a great many,and they can produce many monitoring data for many times in a period time.As a result,the size of the data volume increase sharply,by means of parallel data mining algorithm,realize the rapid analysis of huge amounts of power transformer monitoring data.Spark is a distributed computing framework,with a lightweight rapid processing,compatible with the Hadoop ecosystem,the cost is low,the active learning community support,support for multiple language programming interface etc,in order to handle the vast amounts of transformer monitoring data in parallel,it can provides a new research idea.Common types of transformer faults have been introduced in this paper.Then introduces the traditional and intelligent fault diagnosis methods in detail,and analyzes the advantages and disadvantages of different methods,and proposed power transformer parallel diagnosis and analysis method based on the Spark cloud platform.Choose the Spark machine learning repository naive bayesian method for power transformer fault classify,with the DGA monitoring data as input data,complete fault classification experiments in parallel.The experiment results show that based on the parallel classification method,Spark classification method is superior to single environment in terms of performance.In addition,based on the research of fuzzy clustering algorithm,using the distributed matrix and broadcast variable technologies,completes coding the Spark-FCM fuzzy clustering algorithm in the Spark platform,and extends the Spark machine learning algorithms libraries.And applies the algorithm in the transformer fault clustering,experiments show that the method has good feasibility.