Optimal sensor placement in smart buildings

点击翻译按钮获取中文摘要

According to the United Nations, the building sector has the most potential to achieve cost-effective greenhouse gas emission reductions. Furthermore, buildings need to become more resilient to climate change, as well as more adaptable to the changing commercial requirements. Thanks to the recent development in artificial intelligence and sensing technologies, it is now more affordable to improve operation effectiveness in buildings through software updates instead of costly equipment replacements. Yet there is one major hurdle to the wider adoption of data--driven analytics in buildings - they invariably rely on sensors to capture operation data. Currently, there are no computerized methods for determining where to install sensors, which sensors to select and what to measure. Most of the existing design of sensing capabilities in buildings rely on outdated heuristics and vague rule of thumbs. The aim of this proposal is to develop ways to automatically determine the optimal placement of sensors in building systems, specifically the following three areas: 1) Minimal sensing requirements for virtual metering and key performance indicators (KPI) in building systems. Virtual metering in particular allows building owners use a collection of cheaper sensors to replace otherwise expensive or immeasurable values. 2)Optimal selection and placement of sensors in building HVAC systems to enhance observability and fault isolability. The goal is to determine the optimal amount of sensing required to be able to reliably identify critical faults in HVAC systems, and to steer away from complaint--driven operation to performance- based. 3)Placement of lighting and temperature sensor for flexible space usage. This is to ensure sensor installation within the occupied space can accommodate future layout changes, and provide optimal individual controllability. The main sources of data for this research come from the existing Living lab buildings at Carleton University, as well as data from more than 200 commercial buildings with existing collaborators. The above proposed research will be executed by the applicants and four highly qualified personnel (HQP), including two PhD students, two master's students and two undergraduate students. Trainees will gain significant experience in data--driven analytics such as machine learning, and fundamental engineering theories including optimization, dynamic systems and building physics. These skills are highly sought after by government institutions, industry and education institutes. The research is expected to develop computerized tools for optimal sensor placement in building sensors. This can help lower the barriers to adopting data--driven analytics in buildings. Existing buildings going through a retrofit can easily determine additional sensors needed for data--driven analytics, and new constructions can have these placements calculated during the design stage.