Water-resources management in the Netherlands is undergoing a gradual change of fixed control on the basis of water quantity requirements to integrated and dynamic control, respecting the requirements of all interests present in water systems such as agriculture, nature, urban and industry. Such interests may have different and sometimes conflicting requirements to water availability and water quality, while water resources and their control are limited by capacities of control structures such as pumping stations, inlets, weirs, sluices, etc. Moreover, extreme weather conditions such as heavy rainstorm and lasting droughts, further limit practical control options. A multi-objective optimization problem can be formulated to solve the control problem of a water system under dynamic conditions, however new technologies are nowadays available. Especially Artificial Intelligence approaches are very promising and already show positive results in theory and practise, in particular using machine learning, e.g. Reinforcement Learning. This project focusses on the latter and a polder in the Rijnland Storag Basin will be used as a practical case study, i.e. polder Stein-Oukoop, close to the Reeuwijkse Plassen, where urban, agriculture and nature interests prevail and present-day issues arise: soil subsidence, water quality, biodiversity and sustainability in agriculture. A strategy for the control of the water system has to be developed and tested, building on existing modelling, among which AI models and control tools. The control strategy has to balance the various interests and be based in current and forecasted system behaviour, including weather forecasts and water-systems responses.
Supervision
Are you interested in this assignment? Contact the Master thesis coordinator.
