Upscaling of highly heterogeneous reservoirs is a challenging task due to reservoir simulation of lower resolution scale geological models may present poor quality production indicators. The solution to this type of problem lies in a better calculation of some non-additive geological properties, such as permeability. In recent years, there has been a growth in the development and accuracy of artificial intelligence methods, which can be an important key to the quality of the models, given that AI helps with problems with a high quantity of information. The objective of this work is to evaluate a machine learning-based approach to upscale complex model permeabilities, so that it better captures the global dynamical behavior of different fluids, such as oil, water, and gas. The proposed method uses recent artificial intelligence concepts of machine-learning and optimization to correct the reservoir upscaling in the case of multiple uncertainties resolutions while creating better coarse models whose dynamical behavior honors the fine model. The methodology was adequate and captured the global dynamical behavior of the reservoir. It was able to improve the coarser simulation model results, especially for oil and gas production forecasts, when considering the oil and field results in the AI training.

You can access this article if you purchase or spend a download.