Browsing by Author "Främling, Kary, prof., Aalto University, Department of Computer Science, Finland, Umeå University, Sweden"

Now showing 1 - 1 of 1

Machine Learning Applications for Energy Utilization of Smart Buildings
(2022) Huotari, Matti
School of Science | Doctoral dissertation (article-based)
Energy utilization of smart or intelligent buildings refers to the definition, modeling, and integration of disparate energy elements into coherent energy systems in buildings with the help of artificial intelligence. A core aspect of applications for smart building energy is to address the issues of energy utilization directly while simultaneously taking into account user-comfort, security and malfunctions. Being deployed in increasing numbers in the built environment, these applications are important components of the built environment today. Given the risen number of renewable energy sources together with tightened regulation to energy consumption, the smart building energy applications provide means to combine new technology components together with heterogeneous requirements and goals for energy utilization in buildings. These goals comprise of, for instance, optimal scheduling of energy consumption and production, optimization of costs, integration of renewable energy, user-behavior recognition, and consumer comfort. This research investigates smart building energy applications. This objective is pursued through four research questions which highlight the various aspects of the smart building energy applications: (i) What algorithm to utilize for forecasting the equipment degradation, and what kind of uncertainty is associated with these forecasts for battery packs? (ii) How to build a model in case of gaps or a limited number of observations of interest in data for an air handling unit? (iii) How to involve people in personal environmental comfort decisions for smart building energy applications?, and, finally, (iv) what kind of need is there for smart building energy applications, and which solutions meet these needs? Each of these issues is dealt with using novel techniques, and a related taxonomy was created, as presented in the research publications. The relevance of the proposed solutions is verified with case-studies. Overall, machine learning models solve heterogeneous problems in the field of smart building energy utilization. The results indicate that the proposed solutions can provide answers to a variety of issues regarding building energy management, smart grid, personalization, and maintenance and security.