Physics-Based and Data-Driven Wettability Alteration Model

Abstract

Abstract This paper presents a new wettability alteration model based on surface complexation theory and an extensive experimental dataset. The objective is to provide a general correlation for contact angle calculation that (1) captures the main mechanisms that impact rock-brine-oil wettability and (2) minimizes the number of parameters used to tune with experimental data. We compile a set of 141 zeta-potential and contact-angle measurements from the literature. We study the oil/rock surface-complexation reactions and model the electrostatic behavior of each data point. We develop a new wettability model that estimates the contact angle and consists of five terms based on the Young-Laplace equation. We use the Nelder-Mead optimization algorithm to determine the model-parameter values that produce the best fit of experimental observations. The contact angle estimates produced by our model are also verified against those calculated by Extended-Derjaguin-Landau-Verwey-Overbeekand (EDLVO) theory and are validated using UTCOMP-IPhreeqc to simulate five limestone Amott tests from the literature. The Blind-testing test reveals that our model is predictive of the experimental data (R2 = 0.81, RMSE = 12.5). While reducing the tuning parameters by half, our model is comparable to and–in some cases–even superior to the EDLVO modeling in predicting the contact angle measurements. We argue that EDLVO modeling has 10+ parameters, and the individual errors associated with each parameter could lead to wrong predictions. Amott-test simulations show excellent agreement between the proposed wettability-alteration model and experimental data. The rock's initial wettability was measured to be strongly oil-wet, with a negative Amott index and recovery factor around 5%, corroborating the calculated contact angle of 160 degrees. The recovery factor increases to about 20-35% as the rock becomes more water-wet after interaction with engineered water (contact angle changes to 90-64 degrees). Further analysis indicates the proposed model's capability to capture significant wettability-alteration trends. For example, we report increased water-wetting as brine ionic strength decreases, depicting the low-salinity effect. In addition, our model resulted in better convergence in some of the simulated core floods compared to EDLVO modeling. We conclude that our physics-based and data-driven model is a practical and efficient approach to predict rock-brine-oil wettability.