Research Of Data Mining Method For Public Buildings Energy Consumption Based On Hadoop

The work of energy audit on public buildings is implementing in China. Many provinces and cities have established the large public building energy monitoring system. At present, the number of public buildings being monitored is more than 1000 in Shandong province. Accurate statistical data of building energy consumption has great significance to promote the energy saving in architecture.But the utilization of building energy consumption data is still inefficient. With the increasing amount of building energy consumption data, the analysis of the data is also facing enormous challenges. There is a wealth of information behind these huge amounts of data. It is difficult to discover and summarize the regularities contained in the data using traditional analyzing methods.This paper explores the application of data mining technology based on Hadoop in the analysis of massive building energy consumption data.Hadoop is an open-source distributed computing platform based on Java belonging to the Apache Software Foundation. Users can set up Hadoop cluster with low-cost hardware. Also, users can make full use of the storage and computing capacity of cluster servers to analyze the massive data.On the basis of in-depth analysis and research on the distributed computing platform of Hadoop, in this paper, a new method based on Hadoop for data mining of public buildings energy consumption combining with building information is proposed, to solve the problem of inefficient utilization of building energy consumption data.Firstly, the paper designs the data mining system of public building energy consumption based on Hadoop, and performs designs and illustrations to the basic framework and functional modules.Secondly, on the basis of in-depth study and research on the infrastructure and data reading and writing mechanism of HDFS, the paper writes the Java programs based on MapReduce model, to write XML files of the experimental sample data to HDFS.Then, according to data analysis requirements and the characteristics of experimental data, this paper selects the data mining algorithm. On the basis of in-depth study and research on the infrastructure and tasks operating mechanism of MapReduce, Apriori algorithm and C4.5 algorithm are implemented distributedly using MapReduce programming model. Meanwhile, the paper writes the Java programs of these algorithm.Finally, the Hadoop experimental platform was set up using four computers.The paper takes 200 office buildings in Shandong Province as examples to analysis the data of air conditioning system energy consumption. The experimental conclusions are the influence rules of 7 kinds of building information on air conditioning system energy consumption. Moreover, the experiment obtains the decision tree of air conditioning system energy consumption. According to the decision tree, we can distinguish the energy consumption level of air conditioning system.This paper uses a new method based on Hadoop for data mining of public buildings energy consumption combining with building information, to utilize building energy consumption data efficiently. This method makes up for the defects of previous data analysis methods for massive data processing, such as high cost and low efficiency. The paper explores data analyzing functions of information management system for energy conservation of public buildings in Shandong province. With this method, we can analyze the data of building energy consumption objectively, and provide suggestions for energy saving work. The ideas and methods involved in this paper may be promoted and applied to the data mining on energy consumption of various types of buildings, and provide some references for energy-saving reform of the existing buildings and energy-saving design of new buildings.