The operation of Water Distribution Networks (WDNs) requires methods (i.e. control algorithms) for operating actuators such as pumps, valves, and chlorine injections. Most existing control methods are based on control classic control theory, while only very little work on data-driven control (e.g. reinforcement learning) for WDNs exists.
The aim of potential BA/MA theses or projects is to investigate data-driven control strategies such as reinforcement learning for successfully operating WDNs. Depending on the student’s interest and the required number of credit points, the focus can be on implementing and testing existing reinforcement learning and classic or data-driven control strategies and empirically evaluating those in (simulated) benchmark scenarios, on the creation of new benchmark scenarios, or on investigating (empirically or formally) other relevant aspects such as robustness.
It is also possible to work on this as part of a paid student/research assistant job.
Keywords: Reinforcement Learning, Deep Learning, Critical Infrastructure, Water Distribution Networks
Literature