HVAC performance in commercial buildings is a significant expense, and at the same time it is important for employees health and well-being. Utilization of the office building varies throughout the day, creating opportunities for dynamic optimization.
HVAC systems in office buildings require extensive calibration by human experts. Their performance typically does not improve over time, and may in fact deteriorate as the use of the building evolves.
SentianController can provide effective HVAC control via model-based reinforcement learning. It is capable of optimizing complex combinations of objectives (such comfortable temperature, low VOCs, energy efficiency, etc.). Also, based on learned system dynamics, it timely considers external factors (such as weather) when “reasoning” about the future effects of present decisions. It adapts to changes over time.