Research On Data Mining-based Benchmarking And Evaluation Of Residential Building Energy Consumption

Building energy benchmarking refers to evaluate the energy performance and corresponding energy-saving potential of a given building, thus providing feasible and effective energy-efficient strategies. In the context of the building energy saving and emission reduction, nowadays the research on this field has attracted much attention. Due to the inconsistent characteristics of buildings with different types(such as residential buildings and office buildings) and the complexity of various influencing factors, a key question on building energy benchmarking for a certain type of buildings is how to fully consider the impacts of its influencing factors simultaneously when evaluating energy-use performance. Take residential buildings as an example, a feasible solution can be described as follows: firstly, collecting energy-related data of a number of selected buildings and establishing corresponding database; then, classifying these buildings into different groups based on their influencing factors so that buildings in a same group bears striking similarities; lastly, identifying benchmarks for each group and evaluate the energy-use performance. To date different methods have been proposed for building energy benchmarking. However, these methods hardly can take various influencing factors into consideration simultaneously while the threshold for classification are highly subjective.In this paper, a new methodology for residential building energy benchmarking and energy-use evaluation is proposed based on data mining techniques. Grey relation analysis and cluster analysis are employed to establish a building classification model. Grey relational analysis is used to analyze different influencing factors(i.e. characterized parameters) of total building energy consumption. The grey relational grade is used as the weights of corresponding factors. Based on the weighted parameters, cluster analysis is performed to classify buildings into different groups. Then, accumulative analysis is conducted to identify the benchmarking values for each group. Finally, by comparison with benchmarks and other similar energy-efficient buildings, energy-saving potential and useful recommendations could be provided.To demonstrate the applicability of this methodology, it was applied to the database developed by The Architecture Institute of Japan with the main goal of establishing a benchmarking system. In this system, twelve influencing factors were selected as parameters for cluster analysis and all the buildings were classified into four groups based on these parameters. Corresponding benchmarking values for them were 391MJ/m2, 425MJ/m2, 327MJ/m2 and 390MJ/m2 respectively. To evaluate the energy-saving potential of buildings, a certain building in cluster 1(Building A) is selected as an example. Compared with the benchmarks in this cluster, the energy-saving potential of this building is identified as 157MJ/m2. Furthermore, in order to provide specific energy-saving strategies, it is compared with the energy-efficient building with the most similar characteristics in the same cluster. It is recommended that building owners should give top priority to the following suggestions: firstly, improve the performance of building insulation and air tightness of windows and doors through energy retrofitting; secondly, modify energy-use behavior to reduce energy consumption associated with HVAC and HWS.The proposed method is able to consider different influencing factors simultaneously and to identify the threshold for classification rationally. Therefore, energy-use performance can be assessed more accurately and feasible and effective energy-saving strategies can be provided. This method could be further applied to other types of buildings such as office buildings and hotel buildings. In addition, occupant behavior is widely considered as significant influencing factor of building energy consumption, while its impacts on building energy benchmarks is still unclear. Thus, the main focus of future research should be placed on establishing new building energy benchmarking and energy use evaluation methods with occupant behavior being considered.