Distribution patterns of energy consumed in classified public buildings through the data mining process

Reliable spatio-temporal distribution analysis of building energy consumption is a crucial basis of bottom-up regional energy models, especially when faced with uncertain information at the planning stage. In existing statistical models, the regional consumption levels were mostly identified based on a large amount of samples with multi-dimensional parameters, which are usually not available for developing countries. Pointing to this, the distribution features of regional energy consumption are explored in this paper based on the whole procedure of data mining, which consists of three parts namely pre-processing, information mining, and validation & application. In this process, 212 samples of classified public buildings in Beijing and 66 samples in Hangzhou are included. Firstly, the pre-processing is conducted stepwise aiming at processing the missing data and the abnormal data. Afterwards, the interdisciplinary Lorenz curve is introduced to transfer the scatters into regular curves with satisfied fitting goodness. Thus, empirical formulae are extracted to quantify the nonlinear distribution principles of individual EUIs along with the accumulative building area. Finally, the achieved empirical formulae of different building types are validated, and the application potential of the identified patterns is discussed aiming at the planning stage. Through data mining of the limited datasets, this paper attempts to identify the hidden distribution patterns of regional energy consumption, which enables the regional modeling.