Comparative Analysis of Standalone and Ensemble Machine Learning Models for Enhanced Petroleum Production Prediction

Background and Objective: In this study predictive analysis presents a comprehensive examination of petroleum production forecasting. It focused on predicting oil, gas, and water production using advanced machine learning (ML) techniques. Materials and Methods: Seven standalone models, which include Multiple Linear Regression (MLR), random forest regression (RFR), XGBoost, Support Vector Regression (SVR), decision tree regression (DTR), Artificial Neural Network (ANN), and Rotation Forest (PCA with Random Forest as the base model), were developed and evaluated. Additionally, developed a stacked model that combines Random Forest and XGBoost with Linear Regression as the meta-model. A weighted average ensemble of Random Forest, XGBoost, and artificial neural network was implemented Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and determination coefficient (R²) score are among the evaluation metrics that utilized in measuring the performance of the models. Results: Among the standalone models, RFR achieved the best performance. It outperformed the stacked model. However, the weighted average ensemble outperformed all other models. It achieved an impressive R² score of 0.949 for oil production, 0.948 and 0.968 for gas and water production, and also it achieved least RMSE score. Conclusion: This analysis highlights the effectiveness of ensemble techniques, particularly weighted averaging, in accurately predicting petroleum production. They show a potential for upscaling the decision-making act in the oil and gas industry.