Multi-stage linear programming optimization for pump scheduling

Abstract

This study presents a methodology based on Linear Programming for determining the optimal pump schedule on a 24-hour basis, considering as decision variables the continuous pump flow rates which are subsequently transformed into a discrete schedule. The methodology was applied on a case study derived from the benchmark Anytown network. To evaluate the LP reliability, a comparison was made with solutions generated by a Hybrid Discrete Dynamically Dimensioned Search (HD-DDS) algorithm. The cost associated with the result derived from the LP initial solution was shown to be lower than that obtained with repeated HD-DDS runs with differing random seeds.