Human Building Interaction System Design - Energy Awareness and Learning Approach

Abstract

This thesis presents an outlook to human building interaction. Those interactions happening inside living spaces usually invoke energy consumption. That's why they are important. People are dealing with buildings in order to meet their comfort requirement. The work explores methods giving awareness about energy-related behaviors to people and at the same time an advanced learning system connected to sensors and actuators is designed to learn occupants behaviors inside the buildings. It addresses a bottom-up approach for energy management. Future smart cities will need smart citizens, thus developing an interface connecting humans to their energy usage becomes a necessity. The goal is to give a touch of energy to occupants' daily behaviors and activities. Then, making them aware of their decisions' consequences in terms of energy consumption, its cost and carbon footprint. Moreover, to allow people directly interacting and controlling their living spaces, that means individual contributions to their feeling of comfort. All the work is concentrated on a main concept key: to motivate population modify and optimize their habits concerning energy consumption. It is achieved by a a software package in which user is connected to energy concept and can see his consumption in the living space that he occupies. Such a web-based platform will oer its users the possibility of acting on their living spaces, reacting about their energy-related behaviors and nally interacting with other occupants in order to create a community caring about the energy and its impact on the environment. Therefore, a human building interaction system with energy awareness aspect is designed to keep track of all personal energy related events and its possible features are explained. Data arriving from building's infrastructure has human behaviors and decisions hidden inside. Deriving useful knowledge in the format of buildings' experience is another target of this thesis. The core is an approach to automatically create a knowledge base for smart buildings and let it dynamically evolve based on the behaviors of occupants. Such a knowledge base compromises all life-style experiences and intelligence of a building. Accordingly, management decisions and control can be performed. To achieve the goal, a learning approach is followed to catch knowledge and generate rules from the sensed environmental data and actuators states. Rules are fed to the system as fuzzied events. Applying such a system to maintain the knowledge base, abstracts the complicated coordination between human, infrastructure components and static structures such as building floor plan and architecture. Personalization is a further step to make the ambiance intelligent. Using a prototype what is already possible has been veried. It means such a knowledge base not only can be used as the brain for building to oer person-wise services but an energy related habits prole for the occupant which can accompany him in his movements from a living space to another through smart city. The performance of learning approach has been veried by deploying it on Iris database to make it comparable with other existing methods and its performance analysis with bench marking. The online learning algorithm and decision making has been applied on a demonstration wall as an emulation for a real world single ofifce. Finally only the learning algorithm in offline mode has been applied on data coming from a real office building