Study Of Data Mining Technique Of Analyzing Energy Efficiency For Public Buildings

The public buildings, which are characterized by high density of power consumption, refer to those used for all kinds of public activities besides the house including office buildings, commercial buildings, tourism constructions, science and education buildings, communication buildings and transportation spaces. Researching on the building energy saving especially for those large public buildings is the focus of work in building energy saving in China. During the process of measuring the power consumed by devices in building, massive quantity of real-time energy consumption data were accumulated. Due to their huge quantity and high-dimensional features, abundant knowledge behind them were barely found and concluded by usual conventional analysis methods. However, the data mining technique has powerful data analysis ability and is qualified to discover potential and useful knowledge from massive data. It provides a new way to discover the information and knowledge hidden in massive data.The thesis conducted energy consumption analysis using data mining technique to improve efficiency in making energy-saving decisions Main research contents and methods are as follows.①Quantifying the energy utilization. A concept of energy consumption distribution rate and its calculating formula was proposed to cover the disadvantage of simple judgments on energy utility efficiency.②Establishing clustering mode of public building energy consumption analysis. The data mining application themes are energy consumption prediction, building energy consumption benchmarking and operation optimization while the last one is focused on in the thesis. Clustering subjects were proposed for data mining. The whole process of mining data, which includes data preprocessing, pattern discovery, result interpretation and evaluation, was illustrated. Corresponding solutions were given for problems in each phrase respectively.③Second developing of data mining tool Weka. The original Weka system only integrated several traditional algorithms. The thesis presented a promoted edition of Weka which was second developed by embedding Chameleon algorithm, extending system function and Chinese localized. It can be included that the key point to do second development is to fully understand corresponding data interface and use the classes in Weka. The Chameleon algorithm was integrated into Weka system and was compared with other algorithm in it. The testing result show that chameleon algorithm can obtain better clustering results and are capable to discover any shape and any size of cluster.④A more precise and reasonable building energy consumption clustering mining model was established using data collected from an office building and a commercial building on new Weka system. According to this model, time distribution law can be drawn. The model was used to get three time phrases according to the energy consumption distribution law to reach the goal of saving energy. It also could be used to conducting diagnose on energy saving for corresponding type of buildings.The results show that the method of data mining needs barely too much expertise, avoids complex formulas and focus on dealing with data. Better analysis out comings can be acquired through it. Due to restriction in researching ability and environment, there are many issues still keeping untouched. However, the thesis provides a new method and idea which can be used to for the future research in analyzing building consumption and decision-making on energy saving.