Data-driven Predictive Control Optimization for Natural Ventilation in Buildings

Abstract

Natural ventilation reflects an emphasis on self-control in architecture: it is a crucial aim in the creation of healthy, comfortable indoor environments with low energy consumption. The design of naturally ventilated buildings and their control algorithms signifies the extension of design boundaries to stochastic elements in the natural environment. Unlike the heating, ventilation, and air conditioning (HVAC) system, the driven forces of natural ventilation, such as the wind and the temperature difference, are not reinforced by a mechanical fan. In buildings, the only controllable component of natural ventilation is the operable window. Conventional natural ventilation control strategies lack the capacity fully to use weather information (e.g., outdoor temperature, wind direction and speed) or rely on rule-based control. Therefore, such strategies typically do not fully achieve the goal of high energy performance. This research developed a controllable natural ventilation system (CNVS) to co-optimize the occupant’s thermal comfort, indoor air quality and building energy efficiency, as well as its supportive Internet of things (IoT) research platform. The seasonal CNVS is a 2-layer modular structure. The three CNVS systems (summer, winter, spring/fall) used a combination of a daytime module, a nighttime module, and an occupancy module. The development of the CNVS framework relied on the data-driven predictive control. The evaluation of CNVS was realized in the lab of Harvard HouseZero building.