Model for Optimal Operation of Water Distribution Pumps with Uncertain Demand Patterns

Puneet Khatavkar, Larry Mays

Research output: Contribution to journalArticlepeer-review

18 Scopus citations


An optimization model is presented for pump operation based upon minimizing operation costs and indirectly the maintenance costs of pumps considering uncertainty of specified demand (load) curves. The purpose of this model is to determine pump operation to meet the uncertain demands as well as to satisfy the pressure requirements in the water distribution system. In addition, constraints on the number of pump (‘on-off’) switches are included as a surrogate to indirectly minimizing the maintenance costs. This model is a mixed integer nonlinear programming (MINLP) problem using a chance constraint formulation of the uncertain demand constraint. The optimization model was solved using the LocalSolver option in A Mathematical Programming Language (AMPL). The model was first applied to the operation of an example pumping system for an urban water distribution system (WDS) illustrating a reduction in operation costs using the optimization model. The optimization model with the chance-constraint for demand was applied for a range of demand satisfaction uncertainties. A decrease in the operation costs was observed with an increased uncertainty in demand satisfaction, which shows that the model further optimizes the operations considering the relaxed constraints. Model application could be extended to operations of pumping systems during emergencies and contingencies such as droughts, component failures etc.

Original languageEnglish (US)
Pages (from-to)3867-3880
Number of pages14
JournalWater Resources Management
Issue number12
StatePublished - Sep 1 2017


  • Optimal operation
  • Pumps
  • Uncertain demand
  • Water distribution

ASJC Scopus subject areas

  • Civil and Structural Engineering
  • Water Science and Technology


Dive into the research topics of 'Model for Optimal Operation of Water Distribution Pumps with Uncertain Demand Patterns'. Together they form a unique fingerprint.

Cite this