Research On Fault Diagnosis Model Based On Spark And Data Sharing Among Cloud Platforms

With the continuous advancement of smart grid construction,the number of intelligent primary power equipment and conventional power equipment will increase,and the monitoring data have become increasingly large.Data centralization and data sharing will become the inevitable trend of grid construction,Centralized mass data fault analysis and processing online real-time data will face challenges.The cloud platform dealing with these mass data has the natural advantages to meet these needs.In addition,the performance of deep learning applied to various fields is better,and gradually replaced other machine learning algorithms.Fault diagnosis analysis of power grid equipment using deep self-coding network and data sharing using Alluxio based on cloud platform will be studied in the article.An architecture is designed for fault diagnosis of power equipment monitoring data and data sharing among cloud platforms.Based on the architecture,taking transformer of important device of power grid as an example,the implementation of Distributed Deep Auto-Encoder fault diagnosis model based on Spark is studied.The trained model parameters are prepared by Alluxio for Storm cloud platform,which paves the way for the fast real-time diagnosis of Storm.The experiment first classifies the classic classification data set by using the DDAE model on Spark,and experimental results show that the DDAE model is suitable for classification data.Then,the DDAE model and the Bayesian model are tested by using the transformer DGA data.The experiments through the single and cluster experiments of the DDAE model show that the cluster can accelerate the model training,and the accuracy of experimental results show that the DDAE model has a higher diagnostic accuracy than the Bayesian model in fault diagnosis of transformers.Alluxio links data from HDFS(Hadoop Distributed File System)or hard drives to Alluxio,and Spark and Storm platforms can access data through Alluxio.Experiment on Alluxio show that Spark and Storm not only enable data sharing with Alluxio,but also make it easy to access specific data with Alluxio.