Physics driven AI coreflooding simulator for SCAL data analysis

Two of the most critical properties for multiphase flow in a reservoir are relative permeability (K_r) and capillary pressure (P_c). To determine these parameters, a meticulous interpretation of coreflooding and centrifuge experiments is necessary. In this work, a machine learning technique was incorporated to assist in the determination of these parameters quickly and synchronously. A state of the art framework was developed where a large database of K_r and P_c curves were generated based on existing mathematical models. This database was used to perform thousands of coreflood simulation runs representing drainage steady state experiments. The results obtained from the corefloods including pressure drop and water saturation profile along with other conventional core analysis data were fed as features into the machine learning model. The entire data set was split into 70% for training, 15% for testing, and the remaining 15% for the model validation. The 70% of the training data teaches the model to capture fluid flow behavior inside the core. K-fold cross validation technique was also utilized to increase the accuracy of the model. The trained/tested model was thereby employed to estimate K_r and P_c curves based on available experimental results. The values of the coefficient of determination (R2) was used to assess the accuracy and efficiency of the developed model. The respective cross plots indicate that the model is capable of making accurate predictions with error percentage less than 2% on history matching experimental data. Furthermore, the latter implies that the AI-based model is capable of determining K_r and P_c curves with less effort and better reliability as opposed to the conventional way of creating an entire simulation model. Additionally, the results include multiple options for K_r and P_c curves from which the best solution can be determined using engineering judgement. This is unlike solutions from existing commercial software, which usually provides only a single solution. The model currently focusses on the prediction of K_r and P_c curves for drainage steady state experiments; however, the work can be extended to capture the imbibition cycle as well.