A Reference Architecture for Data-Driven Smart Buildings Using Brick and LBD Ontologies

Abstract

With the increasing adoption of sensors, actors and IoT devices in existing buildings, the real estate sector is becoming increasingly automated. Not only do these devices allow to monitor these buildings (energy use, occupancy, indoor air quality, etc), they also enable modelpredictive control (MPC) through building automation and control systems (BACS). A critical feature to enable these is the metadata associated to data streams obtained from the building. Such metadata allows building operators to assess what these data streams are, what they are measuring and how. This can be achieved using metadata schemes and vocabularies, such as Brick, Haystack, Linked Building Data, Industry Foundation Classes. Merging these model-based metadata schemas (semantics) with data-driven monitoring and control (machine learning) into a functional system architecture is a considerable challenge. In this paper, we review the mentioned technologies and propose a draft reference architecture based on state-of -the-art research. This reference architecture is evaluated using a set of predefined criteria.