Multidimensional Analysis And Diagnosis Of Electric Energy Metering Device Based On Big Data

With the smart grid infrastructure improvement,sensor technology,communication technology and computer technology have been widely used in smart grid,data has been generated fastly in systems of measurement,acquisition and monitoring in power grid.It has become a very important research topic that how to store the data in a scalable way and analyze the potential value of it in a short time.At present,large data processing technology has a variety of models,such as off-line processing,memory computing and so on.Off-line processing technology represented by the Hadoop can realize distributed storage and computing,but it has a certain gap in calculation speed compared with the memory computing.The analysis and fault diagnosis of electric energy metering device information have a higher requirements in timeliness,so it has been a focus of attention of scholars and experts that how to make multidimensional analysis and diagnosis of the metering device big data in a short time.Firstly,the source of the relevant data of the electric energy metering device is introduced in this paper,it shows that the quantity of electric energy metering device,which is represented by smart meter in smart grid is very large,more and more relevant information is collected in the electric information collection system,and it gradually shows the characteristics of big data such as huge amount,kinds of types,and fast growth rate,the timeliness of the data processing requirements are relatively high.Then the advantages of memory computing compared with the off-line analysis processing method represented by Hadoop according to these characteristics and requirements is explained.And then the abnormal types of measuring devices are modeled according to abnormal types of metering devices,and the overall scheme of analysis and diagnosis of abnormal information of metering device is established.The diagnostic procedure and method of several typical abnormal cases are introduced.Then a big data platform is built,and the abnormal characteristic value of the data of the metering device is calculated by Spark SQL and HQL,the analysis of computing result is made.And then the naive Bayesian algorithm is introduced in detail,so fault diagnosis of metering device based on Cluster Advantage can be made.In the end the performance of cluster anomaly characteristic value calculation and the result of the calculation of abnormal characteristic value are tested,several analysis examples are listed.The efficiency and resource occupancy of Spark SQL and HQL in processing data are compared,it verifies the feasibility of using the parallel naive Bayes algorithm to diagnose the anomaly of metering device,the efficiency advantage of memory computing compared with single machine and off-line batch processing is analyzed.Finally,the experimental results show that the cluster has a very good speedup.