Perpustakaan Digital ITB

The need for oil and gas drives the optimization of oil recovery in all reservoirs to the fullest. However, several production constraints resulted in non-optimal oil production from the reservoir, one of the causes was coning. Coning is a production problem in which gas cap gas or bottom water infiltrates the perforation zone in the nearwellbore area and reduces oil production. Coning is a rate-sensitive phenomenon generally associated with high producing rates. This phenomenon occurs when the pressure forces drawing fluids toward the wellbore overcome the natural buoyancy forces that segregate gas and water from oil. There are several strategies to predict the coning problem in terms of critical oil rate with analytical, empirical, and numerical approaches. Generally, a simulation with a numerical approach is used for reservoir modeling. In addition, there is an approach with a statistical method through a proxy model which is considered simpler. A proxy model is a type of machine learning built on large amounts of experimental data. The proxy model is becoming more widely used because it can simplify very complex processes with reasonable accuracy. The proxy model is used in the form of a response surface to speed up interpretation and optimization methods. In this study, by using a proxy approach to the model, the authors overcome the problem coning by sensitivity to perforation and some other effect parameters. The results of this study after testing the case on the validation reservoir with gas and water coning problems, the proxy model can predict the optimum perforation length to get the maximum cumulative oil production at water breakthrough quickly and easily with reliable results.