Resumo:
In a water distribution network, the electricity amount to operate the pumps can achieve
90% of the total electricity consumed. The amount charged for electricity consumption can
differ at each time of the day. Therefore, scheduling the pump’s operation at opportune
times can reduce energy costs. Optimal or near-optimal pump scheduling is not trivial
given the nonlinear constraints of the WDN, which include the pump scheduling problem
in the NP-hard class. The pump scheduling problem consists of obtaining the lowest
operating monetary cost, guaranteeing that water is delivered to all demand points,
and without violating the physical constraints of the water distribution network. This
work proposes heuristic methods based on simulations, combining them in an Iterated
Local Search (ILS) metaheuristic. Computational experiments show that the proposed
approach is promissory, obtaining the best results using the binary representation with
the restrictions presented in this work, and when compared to other pump scheduling
representations, the values obtained for the Vanzyl instance reached solutions that deviate
by only 0.73% of the best-known value and at only 2.02% value for Richmond.
