Optimizing booster stations

Abstract

Booster stations are fluid systems consisting of interconnected components such as pumps, pipes, valves and fittings. One of their main applications is to supply whole buildings or higher floors with drinking water if the supply pressure of the water company is not high enough to guarantee a continuous supply for all consumers. This means that a booster station must increase the pressure of supplied drinking water at a given time-variant flow rate. The consumer's demands must be matched at any time and the system operation is restricted by the general laws of fluid mechanics. A common approach to handle the corresponding optimization problem is to model it as a mixed integer linear program (MILP) and to solve this program using a standard MILP solver. This approach is not suitable for large problem instances with practical relevance as it is not possible to obtain good solutions in reasonable time. Hence, the main obstacle is the optimization speed. In this work, we present an approach to obtain good solutions in reasonable time even for large practical relevant instances. We do this by addressing the problem with heuristics from both the primal and dual side combined with the use of problem specific and technical knowledge. This approach is based on modeling the problem as a MILP as well as a mathematical graph and using both views simultaneously.