Research And Implementation Of Real-time Data Warehouse System For Electrical Energy

Based on one Wenzhou leather company’s energy measuring data real-time monitoringsystem, this thesis analyzes company decision-supported data warehouse logical model andphysical model according to the demand of company executives, and does the specificapplication research in the data extraction,transformation, loading and cleaning techniques,etc.Combined with data mining algorithm, this paper makes application research on powerenergy load prediction strategically and tactically. Finally, real-time data warehouse systemand ETL tools based on this research have been designed and implemented. The maincontents of this thesis are as follows:First of all, the demand on electric power real-time data warehouse system has beenanalyzed, and the system structure of electric power real-time data warehouse has beendesigned. According to business requirement, the two subject structures have been established:power consumption and industry energy consumption. Besides, logical model and physicalmodel structures have been designed for these two subjects.Secondly, two methods on power energy load prediction are designed: strategicprediction and tactical prediction. Strategic prediction is mainly used on medium and longterm power load for company executives while tactical prediction is used on short term powerload for managers in production line. There are two algorithms in strategic prediction: KNNalgorithm and DWKNN algorithm. The experimental data shows that performance ofDWKNN algorithm is much better than KNN algorithm. While tactical prediction mainlypredicts power load by the secondary exponential smoothing method. Since this method haslittle discrepancy, it reaches practical application standards.Thirdly, based on above study, the power energy real-time data warehouse system hasbeen developed and achieved. The system includes: user login and connection configurationmodule, ETL tools module, energy measurement multi-dimensional analysis module,electricity load prediction analysis module, and form and graphic display model. It operates well after testing. The error of predicting results is within an acceptable range for users. Thus,the electric power real-time data warehouse has certain practical application value.