Your interests/skills: Basic knowledge and practical experience with Machine Learning and Deep Learning is required (e.g. having completed the ML and DL lectures). Interests in energy data, data analytics and statistics are helpful.
Details:
Manufacturers are interested in the electricity consumption profiles on the level of individual machines. This enables them to predict their consumption when a machine gets replaced or a new machine is added, to reduce the peak consumption of the factory, and to scale on-site batteries, among other usages.
However, electricity consumption data is usually only available as a single time series for a factory. You will develop methods to disaggregate the factory-wide measurement into time series for individual machines, based on a small dataset with machine-level measurements.
Your tasks are as follows:
- Find and summarize relevant literature about energy data disaggregation.
- Familiarize yourself with the data and task.
- Develop baseline methods and get a feeling for the difficulty of the task.
- Implement the most promising Machine Learning and Deep Learning methods from the literature.
- Do a thorough analysis of the performance of different methods and develop ideas for future improvements.
You will get a paid contract with Fraunhofer ISE. For more information see their job offer.