Abstract: Building energy modelling is necessary for building energy forecasting, management and control. We propose a novel framework using Graph Signal Processing (GSP)-based tech- niques to build the building energy model(s) using smart me- ter data along with occupancy and temperature data. GSP deals with the processing of graph signals, i.e., signals defined on the vertices of a graph. Energy consumption of the build- ing is defined as a graph signal and the edges of the graph are worked out using the available meta-information like oc- cupancy, temperature and other contextual information. The meta information used in our work helps us to identify re- lated nodes in the graph. This concept helps us to use Total Variation Minimization of Graph Signals to model the build- ing energy. The proposed solution is validated on actual data from the office buildings for multiple forecast horizons.
Authors: Yashodhan Reddy N, Naveen Kumar Thokala, Vishnu Brindavanam, Spoorthy Paresh and Girish Chandra M (IIT Hyderabad, TCS Research and Innovation)