In this thesis (submitted for the master of computer science (option AI)), I researched the underlying knowledge (constraints, patterns and rules) behind low voltage electric networks, with the aim of detecting and correcting errors that could come up in the models (digital twin). This thesis was in collaboration with Energyville in another project.

Example errors Example of synthetic data which may contain errors / ambiguities (for example the house to cable connection is ambiguous, or cables get “tangled up”)

Source: https://ai.kuleuven.be/stories/post/2021-11-03-energy-forecasting/img/gis.png

I researched the applicability of methods from the Inductive Logic Programming approach with a focus on explainability and applicability to relational data (such as graphs) to find knowledge from the networks. I also mined association rules pertaining to the general structure of networks (found around 800.000 rules under 12 hours), and evaluated 10 different machine learning models compared to the selected relational rule learner.

• skills: Research, Datamining, Relational Rule Learners
• status: Completed (Summer 2024)