Prediction of Permeability Impairment Due to Scale Deposition in Waterflooding Using Machine Learning Algorithm

Abstract The efficient application of waterflooding as a secondary recovery method in oil and gas reservoirs is often impeded by the complex challenge of scale deposition-induced permeability impairment. The precipitation of mineral scales, such as calcium carbonate and barium sulfate, during waterflooding leads to a reduction in effective permeability, affecting the efficiency of hydrocarbon recovery. Traditional models for predicting permeability impairment lack the ability to capture the intricate relationships among influencing factors. Therefore, this study uses Non-Linear Regression and Regularized Greedy Forests to construct a robust predictive model, identifying key parameters influencing permeability impairment and validating model’s performance against traditional empirical model. The performance of each of this model is compared using performance metrics such as RMSE, MAE, R2. The R2 score of Non-Linear Regression and Regularized Greedy Forests algorithm are 0.911 and 0.957 respectively. This shows that the Regularized Greedy Forest (RGF) model is more accurate for these sets of data and also captures the non-linear relationship between the parameters. In conclusion, the proposed predictive model offers a data-driven solution to forecasting permeability impairment due to scale deposition in waterflooding. This work assists in predicting the extent of formation damage in the near wellbore region of reservoir.