Perpustakaan Digital ITB

In this paper, machine learning is used to help the user selecting infill well-drilling location with much easier, saving time, and effective way compared to what the industry has done with conventional simulators. A Voronoi diagram is involved in the process and the workflow is summarized in a procedure. The procedure is very useful for a field with a lack of data and all types of the reservoir, and in this case, a multi-layer Indonesian reservoir type in Field X is chosen as a sample. Machine learning, part of Artificial Intelligence, has been used in oil and gas industries to maximize their productivity with less effort. As time goes by, several machine learning techniques has been used in many different oil and gas industry sectors. One of the advantages of using machine learning in reservoir engineering applications is they do not require specific physical models but can provide good estimations if there is enough data provided. That is why this study can be useful for all types of the reservoir, in other words, in a reservoir of which exact hydrocarbon production mechanisms are not clearly understood and lack of data. Field X is one of the Indonesian oil fields which reservoirs have unique characteristics, known by having many partial layers in each well. A conventional 3-D earth simulator is needed to calculate the current oil in place per well to choose the new infill well candidates’ location. Nevertheless, the problems are, not every field has its most updated 3-d and many reservoir properties are needed to do the simulations. The method of this study helps to choose the new infill well candidates’ locations by just using machine learning and the Voronoi diagram. It doesn’t need many physical data as the conventional 3-D earth simulator need and saves time. All the workflows are packaged in a procedure, with hope for implementation on the industrial scale.