The Research Of Electricity Demand Prediction Based On Distributed Storage And Computing Platform

With the development and deepgoing of the SG-ERP and smart grid construction, the grid business data is increasing exponentially at a dramatic rate, and data source is more complicated and varied. It is imminent that How to make full use of and do deep analysis on the amount of the diverse data, in order to provide a large number of high value-added services. There, taking 《the notification about carrying the work of pilot study of big data application in 2014 form department of information and communication of the state grid》 as a guide, this article carry out the big data work in HuNan province electric power company, which is based on the key indicators data including the company electricity sales, the whole society electric consumption, various industrial electricity consumption, and the external factors such as seasonal change, natural growth, and then using the technology of big data to set up the electricity forecast model, in order to forcast the electric consumption in the future, and ultimately improve thetimeliness and accuracy of the statistical analysis, at last it can provide decision support for company.In the first, the paper introduces the background and significance of the research, combing the the distributed storage and calculation and the analysis and prediction of power consumption status at home and abroad. At the second time, we research the relevant technology of the distributed storage and computing platforms, such as Hadoop, HDFS, Hive, Ganglia, Sqoop, in order to provide a theoretical basis for further research. In the third time, we design and implement the distributed storage and computing platform Fourthly, with analsising the problems of the electricity consumption forecasting on the current, the paper proposed electricity consumption forecasting every year or month based on adaptive iterative learning model, short-time electricity consumption forecasting used neural network which is optimized by genetic algorithm based MapReduce, which are both based on the distributed storage and computing platforms, and then do experiment to verification. The results show that the two forecasting model proposed in the article is faster and more accurate to predict the power consumption trends for the future. Finally, the electricity consumption forecast system is designed and implemented on the basis of the distributed storage and computing platform, including the system function design, architecture design, core interface implementation and the effect of the system after applicated in Hunan Electric Power Corporation.