Modeling hourly energy consumption in Norwegian buildings

Growing world population, unabated use of fossil fuels, and economies aiming at continuous growth exhaust the planet’s natural resources and add to an augmented greenhouse effect. Besides limiting population growth in less developed regions, reducing per capita energy consumption in more developed regions, substituting fossil and nuclear fuels by renewable energy carriers is considered a major step towards a sustainable development. The integration of renewable energy sources into the energy system can reduce pollutants and greenhouse gas emissions connected to energy conversion processes and ensure energy supply also in a long-term perspective. However, the varying supply of renewable energy supply implies challenges to existing energy systems, where traditionally supply used to follow demand. In order to plan, design, and manage modern energy systems sound estimates on regional energy demand with high temporal and spacial resolutions are needed. Due to the area-wide installation of smart energy meters time series of individual hourly or sub-hourly energy consumption data become available. In combination with cross-sectional information, such as household characteristics or building physics, valuable data sets can be formed, allowing the development of detailed consumption models. In this thesis the key factors for energy consumption in Norwegian buildings are analyzed, and a simple approach for modeling hourly energy consumption in different consumer groups within household and service sector is presented. The models are based on panel data sets consisting of hourly meter data combined with cross-sectional data, weather data, and calendric information. The individual impacts of different heating systems on hourly electricity consumption in households are assessed, yielding for example insights about average reductions in hourly consumption in case air-toair heat pumps or wood stoves are used. Moreover, the impacts of further householdor dwelling-specific variables, such as number of residents or dwelling type, are discussed, and a simple method for disaggregating modeled hourly electricity consumption into a temperature-independent and a temperature-dependent component is applied. Comparing goodness of fit of two regression models based on hourly and daily mean values of local outdoor temperature yields that daily mean values are sufficient for modeling hourly electricity consumption, which facilitates the input data requirements. The modeling approach i