Seminar 1: BrIAS Fellow Prof. Alessandra Parisio
Optimisation-based control of flexible resources in multi-energy networks
Abstract: The growing deployment of distributed energy resources can result in significant environmental and economic benefits but, at the same time, in reduced total system inertia and controllability, hence in new challenges to the power grid operation. Within this context, flexibility (i.e., the ability to adjust to the time-varying grid conditions) plays a crucial role for the transition towards power systems that can efficiently accommodate high shares of renewable energy sources. However, managing flexibility in urban districts and in distribution networks requires control and optimisation tools not yet available. Furthermore, there are several multi-energy systems within a district (i.e., systems with interconnected electricity/heating/gas networks), which currently lack coordination, and which can be regarded as excellent flexibility providers. Novel control strategies and schemes are needed to harness their unique potential. There is still a limited understanding of how to devise effective frameworks for coordinating an arbitrarily large number of flexibility sources. Filling this knowledge gap is essential for the transition to a more sustainable energy grid. In this talk, promising distributed control approaches for coordinating flexible resource, which leverage advanced methods, such as model predictive control and time-varying online optimisation, and data, are explored and illustrative case studies are discussed.
Seminar 2: BrIAS Fellow Prof. Panagiotis Symeonidis
ChatGPT and AI for Medicine
Abstract: In this talk, we will present how LLMs and ChatGPT can make the work of medical doctors more productive and effective. Moreover, we will discuss about methods for finding optimal drug combinations to support the work of medical doctors, by minimizing the unwanted drug side effects (e.g. less drug toxicity) for patients or improving their recovery (e.g. faster healing). In particular, we will present state-of-the-art deep reinforcement learning algorithms for providing personalized medication recommendations. Moreover, we will present graph-based methods, which can find interesting patterns from knowledge graphs related to health, for providing explainable personalized health recommendations. Moreover, we will present a demo application which is based on the Electronic Health Records (EHRs) of patients of a real hospital, where an improved patient’s modelling can help medical doctors to identify early the most critical vital signals (e.g., glucose index, heart rate, etc.) from lab tests related to a patient’s clinical status. Finally, we will demonstrate a proof-of-concept application, which can be used to predict optimal dosing of insulin for patients with diabetes.
