Leveraging data: a new frontier in building modelling and advanced control

Abstract

The ever-increasing availability of data in buildings has sparked a profound transformation across the board in all areas of human activity, in fields as diverse as engineering, entertainment, marketing and medicine. Building performance simulation and building operation are no exception: slowly but steadily, datasets frombuildings are being used for load forecasting, fault detection and diagnosis, the identification of opportunities for energy savings and peak load reduction, optimizing interaction with smart grids and a better understanding of occupant behaviour. International ongoing efforts, such as the work of the IEA EBC Annex 81 ‘Data-Driven Smart Buildings’ efforts, focus on how to better use data to gain insight on building operation and improve their overall performance. While important hurdles have been identified, most notably the need to standardize data labelling and structure in building automation systems, numerous technological advances such as machine learning, in addition to the need to decarbonize the building sector will drive the adoption of data-driven tools over the next decades. In the field of building simulation, the value of data is immense. While building performance simulation rests upon well-understood and rigorous physical principles, thenumerous intervening variables and their interactions make it difficult to assess to what extent the aggregate of these models yields a clear picture of themajor energy flows in abuildingandof its interactionwith thegrid.Data accessibility and treatment will provide an increasingly solid ground for a new paradigm of ‘evidence-based’ building performance simulation, particularly in aspects related to short-term dynamics and building operation. ‘Big Data’, either from a single building or from many buildings, will bridge the gap between the understanding of building physics and mechanical systems, and the educated guesses required in the assumptions made to develop a model. The impact of data is twofold: (a) it will facilitate the task of creating reliable predictive building models with generalization capabilities; (b) it will streamline the implementation of advanced control in a large diversity of building configurations and climatic conditions, with increasingly integrated renewable energy sources such as building-integrated photovoltaics, as well as energy storage systems.