Introducing a dimensionality reduction approach for decommissioning of oil and gas installations

• Ano de defesa: 2019
• Autor(a) principal: Martins, Isabelle Duran
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Universidade Federal do Rio de Janeiro; Brasil; Instituto Alberto Luiz Coimbra de Pós-Graduação e Pesquisa de Engenharia; Programa de Pós-Graduação em Engenharia de Produção; UFRJ
• Palavras-chave em Português: Decommissioning; Decision making; Dimensionality reduction; Machine learning; CNPQ::ENGENHARIAS::ENGENHARIA DE PRODUCAO::ENGENHARIA ECONOMICA
• Link de acesso: http://hdl.handle.net/11422/13281

Decommissioning problems within the oil and gas industry often demand rather involved decision making processes, which may give rise to a large number of criteria since it considers the usually conflicting interests of multiple stakeholders. Moreover, each criterion must be evaluated in connection with each piece of equipment for each available decommissioning alternative. Hence, complex oil and gas fields comprised of a very large number of installations are likely to set up prolonged decommissioning studies. To circumvent this problem, this work proposes the application of feature selection and machine learning supervised techniques to simplify the process. The rationale is to make use of a training set to identify a reduced subset of criteria with significant impact on the selection of the decommissioning alternative. To validate the proposed approach, a dataset was composed based on real-world data from actual sub-sea pipelines through bootstrap techniques. By doing so, one only needs to assess the most significant criteria for all installations without the training set, thus reducing both cost and duration of the decommissioning study as a whole.