Skip to Main content Skip to Navigation
Theses

Optimisation d'alignements d'un réseau de pipelines basée sur les algèbres tropicales et les approches génétiques

Abstract : This thesis addresses operations optimization in an oil seaport with the fundamental purpose of assisting supervision operators. The objective is to provide candidate solutions for pipeline alignment (path) selection and for scheduling of oil transfer operations, as well as maintenance operations, considering that the system has limited and conflicting resources. Informed decision making should consider operations scheduling and alignment selection based on : devices availability, operative capacity of the network, financial aspects (penalties) due to late service, and a predefined maintenance schedule. The optimization problem is addressed by tropical approaches given their potential for yielding concise and intuitive representations when modeling synchronization phenomena. The proposals developed herein start by algebraic mono-objective optimization models. They subsequently become more complex, as new aspects are included, leading to the formulation of hybrid multi-objective optimization models based on artificial intelligence approaches as well as (max,+)-linear system theory. Firstly, a mono-objective optimization model is proposed in (max,+) algebra for penalty minimization. This model integrates different nature phenomena into one single constraint. It is nonlinear, considers predefined alignments for transfer operations and is validated through an optimization solver (LINGO). Secondly, linearization of such model is introduced for prioritization of conflicting operations. Within this context, 2 criteria are addressed: potential penalties for clients and, on the other hand, operations criticality. The nonlinear (max,+) model minimizing penalties is extended in order to consider alignment search and delay minimization for maintenance operations. In order to address the multi-objective nature of the problem, an approach based on genetic algorithms and (max,+)-linear system theory is proposed. Finally, in a more formal framework, a new synchronous product for tropical automata exploiting parallelism phenomena at the earliest is defined in order to minimize the makespan. The proposed models and methods herein have been validated by industrial data gathered from the oil company PDVSA and from the supervision solutions provider Thales Group. The main contributions of this research are, firstly, the application of tropical approaches to this specific optimization problem, yielding concise and potentially linear models. Secondly, the proposal of a hybrid approach based on genetic algorithms and (max,+)-linear systems, which exploits the advantages of the distributed search of artificial intelligence approaches and the conciseness of the models stemming from (max,+) algebra. The final contribution focuses on the definition of the alphabet for a new synchronous product of tropical automata.
Document type :
Theses
Complete list of metadata

Cited literature [76 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-01371572
Contributor : Abes Star :  Contact
Submitted on : Wednesday, May 17, 2017 - 4:05:20 AM
Last modification on : Tuesday, January 5, 2021 - 12:32:03 PM

File

these.pdf
Version validated by the jury (STAR)

Identifiers

  • HAL Id : tel-01371572, version 1

Citation

Karla Rossa Quintero Garcia. Optimisation d'alignements d'un réseau de pipelines basée sur les algèbres tropicales et les approches génétiques. Automatique / Robotique. INSA de Lyon; Universidad de los Andes (Bogotá). Facultad de ingenieria, 2015. Français. ⟨NNT : 2015ISAL0030⟩. ⟨tel-01371572⟩

Share

Metrics

Record views

422

Files downloads

539