Speed of discrete optimization solvers for real time sewer control

Abstract

In the project ‘Integrale sturing afvalwater en oppervlakte water Hoeksche Waard’ (integral management of the drainage and sewer system on the island ‘Hoeksche Waard’) in the southwestern Netherlands a central controller was developed for small sewer systems. The aim was the improvement of water quality through the reduction of combined sewer overflows. Two field tests of the controller went very well. Extension to a larger system resulted in timing problems. This paper investigates these problems. The controller minimizes the deviation from proportional distribution of the volume of sewage over the system. The pumping stations contain several pumps that can be either on or off, so the variables to be optimized do not vary continuously, but are only allowed to have integer values. The computational cost is investigated for different solvers. This cost determines the applicability of a solver for a given system size and time step.

Keywords:

• sewer system
• real time control
• quadratic integer programming
• optimization
• computational compelexity