Augmenting building energy usage models with data segmentation

Energy is the lifeblood of modern civilisation, with buildings and building construction contributing to roughly 40% of the global energy usage and CO2 pollution. Predicting building energy consumption is essential for energy management and conservation; data driven models offer a practical approach to predicting building energy usage. The aim of this paper is to improve the data driven models available to aid facility managers in planning building energy consumption.
In this case study the ‘Clarendon building’ of Teesside University was selected for use in using it’s BMS data (Building Management System) to predict the building’s energy usage. With a particular focus on how data segmentation impacts a model’s accuracy and computational time, in predicting temperature related building energy use. Specifically, the effect of segmenting data to accommodate seasonality, as well as building activity and dormancy periods. With each data segment to be used to train an ANN model (Artificial Neural Network), to address the different patterns and trends present in each period/segment, using ensemble models where data segmentation overlapped.
The potential of these models were compared on the grounds of accuracy to each other, then discussed to identify the various impacts of segmenting the data. This study was performed as part of a larger study, in improving building energy use predictions during the operational period, by incorporating predicted user behaviours.
KEYWORDS: Buildings, Neural networks, Data segmentation, Energy, Prediction.