Area of research:
Diploma & Master Thesis
The position is suitable for part-time employment.
Gaussian processes are a powerful methodology from machine learning for regression problems: given some (often time-dependent) data, find a function that predicts the value of the quantity at a not yet seen time instant. Gaussian processes provide not just a point forecast, but an entire probabilistic forecast in terms of Gaussian normal distributions. The objective of the research project is to take real electrical power consumption data measured at Campus North, and to design a suitable covariance kernel and predict future power consumption. The design of the covariance kernel should include seasonal fluctuations and day/week-day discrepancies. The implementations can be done within the research platform EnergyLab~2.0 at Campus North. Having obtained a Gaussian process for the electrical load(s), this can be employed in optimization-based scheduling of power systems—-which shall be studied if time permits.
If you have a background in a scientific programming language such a Julia, Python, or Matlab, and if you are interested in learning about Gaussian processes, and if you are interested in applying methods to data from the real world, then you are the perfect person.