Banca de DEFESA: JOSE ELOIM SILVA DE MACEDO

Uma banca de DEFESA de DOUTORADO foi cadastrada pelo programa.
STUDENT : JOSE ELOIM SILVA DE MACEDO
DATE: 05/04/2023
TIME: 08:00
LOCAL: Pós-Graduação Engenharia Civil
TITLE:

Development of Optimization Models Based on Evolutionary Techniques for the Design of Water Distribution Networks

KEY WORDS:

Water Distribution Network, Hybrid Algorithm, Metaheuristic Optimization, PSO-RC, Reboot Cycles, H-PSOTS, Computational Cost, Particle Swarm Optimization, Tabu Search

PAGES: 125
BIG AREA: Engenharias
AREA: Engenharia Civil
SUMMARY:

High population growth makes new water distribution network (WDN) projects being designed daily. These networks are an essential part of the water supply systems, continuously and uninterruptedly transporting potable water from reservoirs to the end users, guaranteeing quality, quantity and adequate pressure. However, due to the high implementation and operation costs, and the specific complexity of each network, designers and researchers are developing optimization models that seek the best and most solutions for the sizing, with the lowest possible computational cost. Thus, this thesis presents two new optimization methodologies. The first one is a hybrid model, composed by the Particle Swarm Optimization (PSO) and Tabu Search (TS) algorithms, called H-PSOTS. PSO is based on the dynamics and social behavior of several animal species, while TS is a method that limits repetitive movements on the search. The inclusion of TS in PSO makes the search more efficient, avoiding unnecessary calculations during the solution updates. Both, PSO and TS are evolutionary algorithms, whose optimization technique is metaheuristic, and stand out for presenting robust and logical search strategies. However, the randomness in the formation of the initial solution set of the problem, characteristic of this type of algorithm, can lead to premature convergence in a local optimum and to a reduction in computational efficiency, making the search more exhaustive. To solve this problem, the second proposed model, called PSO-Reboot Cycles or PSO-RC, objective a greater global exploration of the search space with cycles of the algorithm's own reinitialization, without losing information from previous searches, such as the position of the last local optimum found. The models proposed were evaluated on three benchmark problems and on the Panelas-PE water distribution network, and presented better performances than the conventional PSO, resulting in better “optimal” solutions and shorter processing time. Therefore, the proposed methodologies are promising for efficient sizing of water distribution networks.

COMMITTEE MEMBERS:
Externa à Instituição - SILVANETE SEVERINO DA SILVA - UFRPE
Presidente - 2726911 - ANDERSON LUIZ RIBEIRO DE PAIVA
Interno - 2886922 - ARTUR PAIVA COUTINHO
Interna - 2193972 - LEIDJANE MARIA MACIEL DE OLIVEIRA
Externo à Instituição - MARCO AURELIO HOLANDA DE CASTRO - UFC