Big Data Analytics for Refining Processes-EII

• Autor(a) principal: Godinho, Tiago dos Santos Carrasco
• Data de Publicação: 2017
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10362/25382
• topic: EII™; R; manufacturing programs; net consumptions; validation; Domínio/Área Científica::Engenharia e Tecnologia::Engenharia Química

Due to technological development and increased use of renewable energy, the refining sector is becoming increasingly more competitive, therefore it is essential that the refineries reduce their operating costs. Through the EII™2 it is verified that the Sines refinery consumes a greater amount of energy than most of its competitors, when using the same regime and operating conditions.With the aim of minimizing the EII™, a statistical analysis was performed on the refinery data, using agglomeration, dimensionality reduction and elaboration/training of statistical models packages from the tool R. Two general manufacturing programs were identified in the refinery process, grouping four variables associated with the operation process of atmospheric distillation. The net consumptions of 17 procedural units were analyzed through variables such as ambient temperature, quantity/quality of feeding, fouling and type of program. After the analyzes it was concluded that:  The net consumption is not affected by the type of manufacturing program;  The best representation model of net consumption is linear, where the predictor is usually the feeding, the recovered sulfur or the hydrogen produced from the unit. The variables used in the EII™ calculation were then aggregated into multilinear regression models, in which the non-significant variables were identified through the comparison of parameters, such as the normalization coefficient and the p-value. After the optimization of the models it was then performed a validation and training on the same, the same reasoning was used for the models made up of macro variables and variables associated with the individual refinery manufacturing programs. The result was the successful integration of the various models on the daily EII™ chart, as well as the identification of the most influential variables in the EII™, these being related to the consumption of utilities and the operating regimes of AL, atmospheric distillation, HT, HR and HC.