Identifying buildings with rising electricity-consumption and those with high energy-saving potential for government's management by data mining approaches

Abstract

The government in Beijing has established a platform that covered the monthly electricity-use data of 11,370 public buildings, carried out a series of measures to control the quantity of buildings' electricity-consumption and its growth since 2013. Presently, it is trying to find out which buildings that will more probably have a rising trend or higher energy-saving potential than the others, so as to promote the energy-saving momentum and make its management more refine and targeted. Common methods used by the government for these tasks are traditional statistical charts and plots, the analysis results of which always highly rely on the experts' subjective judgment. Moreover, identifying the right objects from such a large database by manpower is also a big challenge. Therefore, this study proposed a series of data-mining methods, including k-means clustering, C4.5 decision tree, 2-dimension scatter diagram and outlier detection, to investigate and model the change patterns as well as compare the levels of the buildings' electricity-consumption. Results indicated that the information wanted by the government was successfully explored from the massive data. Finally, the paper gave some policy proposals on energy-use supervision and data management to help achieve the goal of further energy conservation in public buildings.