(156c) Application of Artificial Intelligence to Predict Surfactant Adsorption

Surfactants play a vital role in chemical enhanced oil recovery (cEOR). The quantity of surfactant loss due to adsorption on a rock directly influences a cEOR project's economics. Therefore, surfactant adsorption quantification is an important area of interest. Surfactant adsorption is influenced by several parameters, which makes its estimation difficult through analytical modeling. In this paper, a real experimental data set was used to predict surfactant adsorption as a function of minerals’ percentage, surfactant equilibrium concentration, and maximum adsorption capacity. Several pure minerals were used to find the static adsorption of a Gemini surfactant using high-performance liquid chromatography (HPLC).

Artificial neural networks (ANN), eXtreme Gradient Boosting (XGBoost), and Random Forest (RF) were applied to build a digital model to estimate surfactant adsorption based on real static adsorption data. The data was divided into an 80:20 ratio for training and testing, respectively. The coefficient of determination (R2) and root mean squared error (RMSE) were used to find the optimal results after sensitivity analysis of hyperparameters. ANN outperformed XGBoost and RF. Feed-forward neural networks consisting of a single hidden layer with four hidden layer neurons gave the best results. The trained ANN model showed good agreement with the unseen data. This research reports a new digital model to predict surfactant adsorption using ANN as a function of surfactant concentration, maximum adsorption capacity, and mineral composition. The newly developed ANN model will save a lot of time in running tedious experiments and will provide a good quick estimate of surfactant adsorption.