Energy disaggregation, or Non-Intrusive Load Monitoring (NILM), is a technique that predicts the consumption levels of individual appliances from only the main signal in the building. Various methods have been proposed to solve this problem, including sparse coding (SC), which offers great advantages due to its ability to capture complex patterns in data. However, a challenging aspect of NILM is that data containing appliance-level information is scarce.  Moreover, the houses that the models are tested on might be from a different population than the training data, thus resulting in a domain shift. Therefore, we need to develop approaches that are adapted to training data scarcity through the use of transfer learning (TL), also known as domain adaptation. In this research work, we explore domain adaptation approaches on SC models with the aim of discriminative energy disaggregation (DD). We compare 4 methods that employ TL, two of which are deep architectures, with 4 methods that do not employ it. In the second part of this thesis, we explore constraining NILM domain adaptation to a privacy-preserving Federated Learning framework. In this case, the NILM models are being trained in a framework that does not allow data to be shown to any model outside of the building's domain. This allows us to experiment with distributed methods in a more realistic setting, where user data is omitted from any third party. For this task, we propose 4 weighted federated domain adaptation methods. We also experiment with weighting methods that further protect the privacy of the user, resulting in a total of 12 approaches that we compared.