Machine Learning for Event Detection in Cyber-Physical Energy Systems
This project investigates machine learning methods for event detection in cyber-physical energy systems, focusing on the distinction between normal operation, physical disturbances, and anomalous system behavior. The work is based on reproducible, literature-driven implementation of established approaches.
An initial baseline is derived from existing work, including data processing, feature representation, and classifier design. The implementation is systematically evaluated using publicly available datasets, with emphasis on reproducibility, robustness, and consistent performance assessment.
Subsequent analysis focuses on model interpretability and the identification of relevant decision factors. Depending on progress, extensions towards improved robustness, alternative model architectures, or enhanced evaluation protocols are considered.
The project emphasizes structured experimentation and reproducible machine learning workflows in the context of safety-critical cyber-physical systems.