Multi-objective optimization of the management of a waterworks using an integrated well field model

Annette Kirstine Hansen, Peter Bauer-Gottwein, Dan Rosbjerg, Henrik Madsen, Anne Katrine V. Falk

Research output: Contribution to journalJournal articleResearchpeer-review

Abstract

This study uses multi-objective optimization of an integrated well field model to improve the management of a waterworks. The well field model, called WELLNES (WELL field Numerical Engine Shell) is a dynamic coupling of a groundwater model, a pipe network model, and a well model. WELLNES is capable of predicting the water level and the energy consumption of the individual production wells. The model has been applied to Søndersø waterworks in Denmark, where it predicts the energy consumption within 1.8% of the observed. The objectives of the optimization problem are to minimize the specific energy of the waterworks and to avoid inflow of contaminated water from a nearby contaminated site. The decision variables are the pump status (on/off), and the constraint is that the waterworks has to provide a certain amount of drinking water. The advantage of multiobjective optimization is that the Pareto curve provides the decision-makers with compromise solutions between the two competing objectives. In the test case the Pareto optimal solutions are compared with an exhaustive benchmark solution. It is shown that the energy consumption can be reduced by 4% by changing the pumping configuration without violating the protection against contamination. © 2012 IWA Publishing.
Original languageEnglish
JournalHydrology Research
Volume43
Issue number4
Pages (from-to)430-444
ISSN1998-9563
DOIs
Publication statusPublished - 2012

Keywords

  • Electric load forecasting
  • Energy utilization
  • Groundwater
  • Multiobjective optimization
  • Pumps
  • Water levels
  • Water management
  • Water supply
  • Waterworks

Fingerprint Dive into the research topics of 'Multi-objective optimization of the management of a waterworks using an integrated well field model'. Together they form a unique fingerprint.

Cite this