Motivation
- Component-level measurements are often unavailable. NILM techniques offer a cost-effective alternative but have primarily been developed for residential settings. (See, e.g. https://link.springer.com/article/10.1186/s42162-022-00230-7?)
Objectives
- Investigate NILM for industrial energy data
- Compare novel generative approaches with a conventional baseline (CNN-, LSTM-, ...)
- Evaluate reconstruction quality for downstream applications
Research Question
- Which architectures transfer best to industrial environments?
Deliverables
- Benchmark comparison
- Open evaluation pipeline
The thesis can be written in German or English. The scope can be tuned based on the personal interests aswell as the type of the thesis. A good to strong knowledge in ML-/AI-methods is recommended to get started quickly.
In case you are interested, please send an inquiry via email. Ideally, attach some information about your person (e.g. CV, transcript of records, ...).
Henning Hupfeld
h.hupfeld@tu-braunschweig.de
Art:
- Bachelorarbeit
- Studienarbeit(Master)
- Masterarbeit
Fachrichtungen: Maschinenbau und angrenzende Fachrichtungen
Begin der Arbeit: zeitnah
Zuletzt geändert: 07.07.2026
