Prediction of Surfactant Performance from Surfactant Huff-Puff in Carbonate Reservoirs using a Data-Driven Approach and Desktop Application Development

Surfactants have been widely used to alter the wettability of carbonate rocks from oil-wetness to water-wetness and enhanced oil recovery. One of primary methods implemented in filed applications for enhanced oil recovery is surfactant huff-puff. Currently, a large number of surfactant treatments are conducted in laboratories prior to field applications to optimize the design of surfactant huff-puff, test the performance of surfactants, and minimize failure risk. This process is time-consuming since a treatment could last from several days to more than 300 days. Moreover, a fraction of treatments could not improve oil recovery as reported in the literature. In this paper, we provide a machine learning based solution to improve this process. A comprehensive dataset that systematically compiles project data on surfactant treatments in carbonate reservoirs is constructed. Based on this dataset, machine learning models are developed to forecast incremental oil recovery resulting from surfactant treatments. Random forest model presents the best performance. This research could predict surfactant performance before a surfactant treatment is conducted, which could fasten laboratory investigation, save time and cost. Furthermore, an adaptive, offline, and friendly graphical user interface is designed to enable data analysis and assist in decision-making. With desktop application, it is easy to conduct data analytics and could be accessible for most engineers.