Perpustakaan Digital ITB

Many of the old large fields in Indonesia have experienced a decline in oil production due to the nature of the field which has decreased pressure levels and the number of new field explorations is lacking. Meanwhile, Indonesia's demand for crude oil continues to increase. This problem makes it necessary to have a method that can increase the level of production in the old big field. The addition of production wells in old fields can be one solution. However, the selection of the new well location requires a large investment, the absence of the latest static model is a problem. Toth et al have developed a formula that can determine the water saturation of a well only with the production data of the well in the oil-water field. This is accomplished using the convenient interpretation formulae derived for radial unsteady state two-phase immiscible fluid displacement in the near wellbore formation. This study was conducted as a validation of whether the built equation can be used. In addition, the well saturation data that has been obtained will be used as input for machine learning using the KNN method, to obtain a map of field water saturation. This process begins with processing the production data of each well along with field geological data to build a trendline. After that, the trendline owned by each well will be seen its R-squared value. If a sum has an R-squared of less than 0.7, then the trendline will be divided into several sectors so that it has an R-squared value of more than 0.7. Based on the trendline equation that was built, it will be the input for the equation to find water saturation based on the equation built by Toth et al. The water saturation value obtained from the equation will be compared with the water saturation value obtained from the reservoir simulation to see the difference. Then the saturation value of the well water will be used as input for machine learning with the KNN method to get a map of field water saturation. The results of this study indicate that of the two fields used as case studies that produce for one year, Field "Z" which has 25 wells, there are only 2 wells that have a saturation difference of more than 0%. As for Field "S" which has 9 wells, there is no saturation of well water which has a difference of more than 9%. As well as these two fields, a saturation map of the field has been formed using the machine learning KNN method. Keywords: Near-Wellbore Water Saturation, K-Nearest Neighbors, Field Water Saturation Map