Methodology for robust production optimization for the management and redevelopment of petroleum reservoirs using machine-learning techniques

Abstract

Reservoir development and management processes are usually computational and timedemanding, requiring an efficient and practical methodology application. The Closed-LoopReservoir Development and Management (CLRDM) proposed by Schiozer et al. (2019) can be apromising methodology for highly complex reservoirs in different field stages, as they describe12 steps to assist the decision-making process. However, the integration of uncertainty inproduction optimization processes of naturally fractured reservoirs can be an extra challenge instep 10 application of the CLRDM. Thus, the use of optimization methods and machine-learning(ML) techniques could accelerate the process and consider all relevant uncertain data. The presentproject focuses on step 10 of the CLRDM process with the main objective of developing amethodology for robust production optimization, focusing on long-term control rules ofequipment over time (G2L) and redevelopment (or revitalization, G3) variables. The applicationwill be on the benchmark UNISIM-II-D-CO (adapted from Correia et al., 2015), a light-oilfractured carbonate reservoir similar to those found in the Brazilian pre-salt reservoir (AU)