Control system for smart (context-aware) and energy-efficient indoor cooling
Background of Research
The need for new buildings to fit within the architecture of “SMART CITY information systems” is becoming a matter of urgency towards achieving energy efficient and zero carbon buildings at the urban level. This research develops a self-learning algorithmic control system that smartly learns from sensory data in a building to achieve energy efficient indoor cooling. The architecture of this system is extensible and adaptable on the smart grid towards urban demand response.
Minimizing the energy redundancies associated with indoor cooling by smartly utilizing building information within deep reinforcement learning HVAC control systems.
1. To assess the level of semantic enrichment that knowledge graphs avail to a building energy model while encapsulating concepts about the heating and cooling load of an office building.
2. To identify the potential that an RDF knowledge graph provides in achieving end-to-end learning within the deep reinforcement learning architecture of an HVAC controlling agent towards more adaptive and optimal control of the cooling load in an office building.
3. To examine and quantify the cooling energy saving potential within a typical office building achieved by the aforementioned self-learning HVAC control system while comparing the results with conventional rule-based systems.
Expected to be completed by second quarter of 2021.
Proposed Integration with ShareBIM
1. A DRL algorithm capable of developing data driven optimal HVAC control policies that minimize unnecessary cooling load in office buildings.
2. A better understanding of how to achieve the balance between indoor thermal comfort, air quality and reduced energy use.
Providing building owners smart options in the way they use their electricity for indoor cooling with possibilities to even fit their buildings into the smart grid towards achieving energy efficiency at the urban level