Distributed model predictive control for flood prevention in water networks

Staff Mentor:

prof.dr.ir. B. De Schutter (Bart)

Other Mentor(s):

A. Sadowska


Distributed and large-scale systems; Multi-agent systems; Distributed control; Optimal and model predictive control


The importance of an efficient and reliable flood and water
management system is increasing, due to among others
higher sea levels, more heavy rain during the spring
season, and possibly also drier summers.
Water management is distributed among many local
bodies. Local control actions include activation of pumps
or locks, filling or draining of water reservoirs, and
opening of emergency water storage areas. Too high water
levels (flooding) and too low water levels (for agriculture
and irrigation) should be avoided, while minimizing the
cost of the local actions.

Local water management bodies usually only control
water levels in a relatively small region. However, the
evolution of the water levels is influenced by what
happens over a much larger region, often extending far
beyond the neighborhood of the given region. The
currently uncoordinated and localized control results in
suboptimal overal system performance.

By cooperating and by coordinating the local
water management actions, and by also taking
into account predictions or forecasts of future
rain fall, future droughts, and the future arrival
of increased water flow via rivers, etc. (using
various weather and hydrological sensors and
prediction models) a more efficient flood and
water management can be obtained with less
risks and less costs. As some of the local
requirements may sometimes be conflicting, a
multi-constraint and multi-objective
coordination and control task has to be solved.

Therefore, we are currently developing novel
intelligent multi-agent model-based predictive
control approaches for flood and water
management. These approaches will guarantee
the basic requirements and service levels to
perform adequate flood and water

© Copyright Delft Center for Systems and Control, Delft University of Technology, 2017.