Julius Aka

Brief description of my doctoral project: AI-Based Surrogate Models for the Energy Transition of the Future

Efficiently shaping the energy transition requires complex simulations of energy systems – from power grids and heat generation to power plants. However, conventional physical models are often too computationally intensive for real-time optimization. My doctoral project addresses this challenge by developing AI-supported surrogate models. These so-called “Balanced Neural ODEs” combine neural ordinary differential equations with model order reduction via variational autoencoders. In close cooperation with industrial partners, I apply this method to optimize the control of heat pumps and power plants. My approach enables the creation of faster surrogate models in a fraction of the time previously required. This makes efficient operational optimization economically viable even for smaller systems, such as residential buildings. Ultimately, my goal is to sustainably bridge the gap between top-tier academic research and industrial application.