Research On The Classification Of Key Clients’ Electricity Consumption Time Series Data Based On Hadoop

With the rapid development of domestic economy, the demands for electricity of everyindustry increase at the same time. For now, Smart Grid has become the center of powersystem, which supports the development of a new generation of electricity production andmanagement, and also automates the information collection. The fast development of SmartGrid has brought massive data for Power Grid Corporation. However, the traditional relationaldatabase is not capable to deal with massive data, and distributed file system becomes one ofthe directions of the future development for data technique in Power Grid Corporation.Besides, simply storing the data fails to promote the development of the corporation.Therefore, people are faced with the problem of how to turn the data into benefits, whichemerges an urgent issue for Power Grid Corporation.Key clients refer to the ones with demands for substantial amount of electricity. Most ofthem are industrial clients or those with special units. Generally, the economic benefits ofpower enterprises are coming from the key clients. Classification of key clients is helpful fortheir management. Based on the classification result, Power Grid Corp can providecustomized electricity supply service, make differentiated electricity marketing and promoteeconomic profits.Key clients’ electricity consumption data shows apparent temporal pattern. With regardto this characteristic, we use time series classification algorithms to classify data. So, in thispaper, the research working mainly includes:(1) We categorize the representative and novel time series classification algorithms intothree algorithms, namely model based, distance based and sequence features basedalgorithms;(2) We set up a Hadoop cluster for our time series classification experiment, anddescribe how to implement all kinds of algorithms with Map Reduce framework;(3) We analyze the effect of Hadoop cluster for the speed of data process and discuss thespeed and accuracy of all the algorithms.The research of this paper focuses on time series classification in data mining anddistributed data process technology. So, it not only presents a data mining application forPower Grid Corp, but also conforms to the development directions of data technology.