Within IRDL; LINEACT, Chinmayi Kanthila has developed methods for detecting and predicting building occupancy.
Chinmayi Kanthila, chercheur·e au sein de IRDL; LINEACT (Institut National Polytechnique de Bretagne).
Thèse soutenue en 2024 à l'école doctorale École doctorale Sciences Pour l'Ingénieur (Lorient ; 2022-....).
Référencée dans le réseau ABES/STAR, cette thèse répond aux critères de rigueur de l'enseignement supérieur français.
The research conducted by Chinmayi Kanthila at the National Polytechnic Institute of Brittany addresses a fundamental issue: the optimization of building occupancy. This topic is not merely technical; it touches on major societal and environmental challenges. Indeed, effective occupancy management is crucial for reducing energy consumption and improving user comfort. In a world where resources are becoming increasingly limited, it is imperative to rethink how we use the spaces we occupy.
By integrating machine learning techniques, particularly LSTM (Long Short-Term Memory) and Bi-LSTM (Bidirectional LSTM) neural networks, this thesis proposes an innovative approach to modeling and predicting occupancy behaviors. These methods, which allow for the processing of complex time series, are particularly suited to the fluctuating dynamics of building occupancy. For example, in office environments, occupancy variations can be influenced by factors such as work schedules, special events, or even weather conditions.
The energy consumption of buildings accounts for approximately 40% of global consumption, a figure that underscores the urgency to act. This observation is particularly relevant in the current context of climate change, where every kilowatt-hour saved can help reduce our carbon footprint. The ability to accurately predict occupancy not only optimizes energy management but also enhances the safety and comfort of occupants. Indeed, precise occupancy counting can influence key decisions such as energy load management and the control of heating, ventilation, and air conditioning (HVAC) systems.
Consider the example of an office building equipped with smart sensors. With accurate occupancy modeling, it becomes possible to automatically adjust temperature and ventilation based on the number of people present in the premises. This not only reduces energy consumption but also contributes to creating a more pleasant work environment, which can increase employee productivity.
Preliminary results from this research show a significant improvement in the accuracy of occupancy predictions. This paves the way for smarter buildings capable of adapting to the real needs of users. The recommendations made by the author include integrating predictive models into public policies, as well as investing in advanced sensor technologies.
These measures could transform the way we design and manage our living and working spaces. By placing occupancy at the heart of energy management strategies, it is possible to reduce energy consumption while improving occupant comfort. This could also have significant economic implications, reducing building operating costs and increasing their value in the real estate market.
Kanthila's research is therefore a call to action for decision-makers, investors, and stakeholders in the construction sector. In West Africa, for example, where urban growth is rapid and infrastructure struggles to keep up, the adoption of these technologies could play a crucial role in creating sustainable cities. Occupancy-centered buildings could become the norm, thus transforming our approach to energy efficiency.
In conclusion, this thesis is not limited to a simple academic study but proposes concrete solutions for a more sustainable future. Occupancy-centered buildings are not just a trend but a necessity in the face of current environmental challenges. By integrating data-driven approaches into building management, we can hope for a future where energy efficiency and user comfort go hand in hand, creating living and working spaces that meet the needs of our time.
Données clés
- 40% : part de la consommation énergétique mondiale liée aux bâtiments
Accéder à l'étude complète
Sources et accès
Chinmayi Kanthila. Towards occupancy-centric buildings : detection and prediction using data-driven methods. Other. Université de Bretagne occidentale - Brest, 2024. English. ⟨NNT : 2024BRES0048⟩. ⟨tel-05273953⟩