The increasing use of renewable energy sources presents new challenges for modern energy grids. Unlike conventional generation technologies, renewable sources depend on environmental factors that determine their output. To address this variability, demand response programs shift the responsibility for balancing energy production and consumption to individual consumers. In these programs, consumers are incentivized to adjust their energy consumption to align with the available supply, for example through dynamic prices reflecting the available energy. However, manually adapting to such demands is tedious and time-consuming, highlighting the need for automated solutions. The growing adoption of the Internet of Things (IoT) provides a promising pathway for enabling smart energy management and supports such solutions.

This requires a framework that can digitally represent real-world environments, simulate behavior based on relevant information and optimize energy usage accordingly. Since heating and cooling in buildings account for a significant portion of total energy consumption, managing these systems represents a particularly effective entry point for smart energy optimization. To this end, this thesis proposes a framework that uses digital twins to model a smart homes heating system. This enables the prediction of future behavior and optimization of energy consumption in response to grid demands. We develop a context model for the digital twin, identify the components relevant to heating energy systems and define a data representation that facilitates the use by machine learning algorithms for prediction. We implemented our framework in a simple setup and found that the collected data with the identified attributes looks promising for predicting future states.