2026

Student: Katharine Buntrock
Supervisor: Tobias Gebhard
Time period: 01/05/2026 - 07/06/2026
Type: Master Thesis


With the rapid electrification of mobility and heating, as well as the rise of solar PV and flexible demand, distribution grids face unprecedented challenges. The increasing energy demand often requires expansion of infrastructure, which is costly. Existing design practices are based on simple, outdated rules, often leading to over-dimensioning of capacities because reliability margins cannot be quantified sufficiently. First, they usually treat power demand as a fixed maximum value, ignoring the inherently probabilistic nature of electric loads. Second, the type of consumers, their individual consumption patterns (e.g. daily/seasonal), and load correlations with other consumers are neglected, but can have a significant impact on the maximum load. For a successful and cost-efficient energy transition, new data-driven approaches for grid planning and optimization are needed.
This thesis aims to change the way, low-voltage (LV) grids are designed by developing a data-driven, probabilistic, correlation-aware methodology. An optimization problem for capacity design, topological transformer placement, and switch placement/configuration is defined and analyzed. The approach is based on multivariate statistical modeling (e.g. normal distribution). To test and evaluate the method, a data analysis of electricity usage patterns from heterogeneous consumers (e.g. residential, commercial, retail, etc) is carried out.
  • Literature review of current practices and methods for LV grid planning
  • Research for electricity demand datasets of small public/commercial buildings (e.g. shops, retail, hotel, bakery etc)
  • Analyze consumer demand correlations of power demand time series data
  • Develop a probabilistic methodology for capacity design, optimal transformer placement, and/or consumer partition by considering the correlations
  • Implement and test the algorithm and evaluate the performance by comparing it with traditional, deterministic approaches

Requirements:
  • Interest in optimization and statistical modeling
  • Basic experience with programming and data analysis (e.g. Python)
  • Attendance in lecture “Data-driven Modeling / datengetriebene Modellierung (Machine Learning)” helpful
  • Attendance in lecture “Energy Management & Optimization” helpful
  • If the thesis is done as B.Sc, very good grades and self-organized acquisition of the prerequisites are expected