Prediction of Water Saturation from Well Log Data by Machine Learning Algorithms: Boosting and Super Learner

Abstract

Intelligent predictive methods have the power to reliably estimate water saturation (Sw) compared to conventional experimental methods commonly performed by petrphysicists. However, due to nonlinearity and uncertainty in the data set, the prediction might not be accurate. There exist new machine learning (ML) algorithms such as gradient boosting techniques that have shown significant success in other disciplines yet have not been examined for Sw prediction or other reservoir or rock properties in the petroleum industry. To bridge the literature gap, in this study, for the first time, a total of five ML code programs that belong to the family of Super Learner along with boosting algorithms: XGBoost, LightGBM, CatBoost, AdaBoost, are developed to predict water saturation without relying on the resistivity log data. This is important since conventional methods of water saturation prediction that rely on resistivity log can become problematic in particular formations such as shale or tight carbonates. Thus, to do so, two datasets were constructed by collecting several types of well logs (Gamma, density, neutron, sonic, PEF, and without PEF) to evaluate the robustness and accuracy of the models by comparing the results with laboratory-measured data. It was found that Super Learner and XGBoost produced the highest accurate output (R2: 0.999 and 0.993, respectively), and with considerable distance, Catboost and LightGBM were ranked third and fourth, respectively. Ultimately, both XGBoost and Super Learner produced negligible errors but the latest is considered as the best amongst all.

Petrophysical Regression regarding Porosity, Permeability, and Water Saturation Driven by Logging-Based Ensemble and Transfer Learnings: A Case Study of Sandy-Mud Reservoirs

“…A strong learning tool is made up of several CARTs, or more technically, several week learning tools [36]. Then, according the computing theory of EL, the modeling implemented by LightGBM can be expressed as [36][37][38]…”

“…Transfer Learning. LightGBM or broadly EL will cause an overfitting or an underfitting when dealing with a small-volumetric dataset and preferentially encounters the underfitting [37,38]. Then, in practical case, there exists a new challenge for the prediction of LightGBM, which should be seriously considered.…”

“…2075 samples are chosen randomly to construct the learning dataset, and the rest is employed as a test dataset. A suggested initial setting of LightGBM is shown in Table 4 [36][37][38]. To stress the optimizing effect of Bayes, RS and particle swarm optimization (PSO) which is a representative of SI are adopted as competitive optimizers.…”

“…Zhou et al [37] employed LightGBM to predict permeability based on a feature selection of well logs, and in accordance with the analysis of the obtained results, ensured LightGBM is a "sharp tool' for the petrophysical prediction. Hadavimoghaddam et al [38] studied an automatic regression for C t = αS n w + γS p w + βS q w C t : conductivity of true formation; α, n: coefficient and power of water saturation term; γ, p: coefficient and power of cross term; β, q: coefficient and power of shale term water saturation, and through a comparison demonstrated, LightGBM is a better selection.…”

Effect of Different Water Content and Catalyst on the Performance of Heavy Oil Oxidation in Porous Media for In Situ Upgrading

“…Viscosity reduction of heavy oil during recovery is the key aspect for the success of the recovery method. − The traditional and existing prediction skills are restricted in terms of precision. Viscosity values are generally determined by laboratory measurements in petroleum engineering applications.…”