The i-SEWER project will develop a methodology for setting up a scalable, autonomous and AI-driven real-time control system for the reduction of overflow discharges from combined sewer systems. Development and implementation of procedures for automatic error recognition in sewer process data, based on Deep Learning algorithms, will ensure that control will be based on reliable system state information. Development of procedures to derive highly performant surrogate models, using Deep Learning combined with hydrodynamic sewer network models, will facilitate the setup of model-based predictive control systems.
Such surrogate models will be integrated with a simulation framework for model-based predictive control and their performance evaluated for example systems. Furthermore, the potential of such model-based predictive control will be evaluated with regard to technical issues and to their sustainability.
ifak will assist the learning process of such surrogate models, developed by the project partners, by provision of dedicated modules within a simulation platform. Furthermore, evaluation of such model-based predictive control based on these surrogate models, including an assessment with regard to sustainability criteria, will form part of ifak’s work packages within the i-SEWER project.
Research and development on these methods will commence using a benchmark example from the literature, but then focus on a real system (sewer network of the city of Freiburg) and subsequently be evaluated from a practical point of view.