Modelo de otimização multiobjetivo baseado em algoritmo Shuffled Frog Leaping para transporte de produtos em redes de dutos

Abstract

The development of model to support pipeline network operation management is an optimization problem which involves complex operational constraints. The product transport through pipelines proves reliable and economical, especially for large volumes. However, the high occupancy rate of the distribution networks and the amount of different products should be transported under different operating conditions lead to complex operational scenarios. An efficiency improvement of products transport through pipeline networks can be obtained by a better allocation of available resources. However that is a hard solution combinatorial problem with multiobjective optimization characteristics. An alternative to efficient solve this type of problem is the use of metaheuristics such Multiobjective Evolutionary Algorithms~(MOEA). MOEA uses a population of solutions in its search, and multiple Pareto-optimal solutions can, in principle, be found in one single run. This work aims to develop a model of multi-criterion optimization applied to scheduling operations in a real-world pipeline network in the oil industry. We use a metaheuristic optimization method inspired from the memetic evolution of a group of frogs when seeking for food: SFLA~(Shuffled Frog Leaping Algorithm). The results obtained from the simulations are compared to an algorithm well known in the literature: genetic algorithm~(GA). Moreover, this works then introduces a new approach of the original shuffled frog leaping algorithm to create a modified form of the algorithm: the Modified Shuffled frog-leaping Pareto Approach~(MSFLPA). The main goal of MSFLPA is to represent and recover the entire Pareto front to a modeled problem, moreover an efficient and competitive algorithm to solve multi-objective scheduling problems with more than two conflicting objectives. This new approach combines the use of a small population and an archiving strategy with a procedure to restart the population using two auxiliary memories to store nondominated solutions (Pareto set) found during population evolution. To validate the performance and efficiency of the proposed MSFLPA in spread Pareto front, five Zitzler-Deb-Thiele functions are examined and compared against two well-known multi-objective genetic algorithms: NSGA-II and SPEA2. The numerical experiments indicate that MSFLPA yields spread solutions and converges to the true Pareto front and it is verified to be efficient and competitive for solving multi-objective problem. After this validation, the MSFLPA is used to optimize the allocation of the resources and to solve the scheduling problem of a real world pipeline network and if compared with NSGA-II and microGA, MSFLPA is verified to be a new effective alternative for solving of multi-objective problems with more than two objectives as it is the case of the pipeline scheduling problems.