Lean Six Sigma & Machine Learning for Optimization: Predicting Heavy Oil Viscosity with Machine Learning to Optimize Steam Injection and Reduce Energy Waste - Softcover

Inhaltsangabe

Heavy oil exploitation presents significant energy and operational challenges due to high viscosity, which directly impacts the efficiency of thermal recovery processes. In Madagascar's Tsimiroro field, accurate viscosity prediction is critical for optimizing steam injection and mitigating excessive energy waste. This study adopts the Lean Six Sigma methodology as a structured framework to eliminate operational Mudas (wastes) related to steam overconsumption. A machine learning-based predictive framework was developed, comparing six distinct architectures: Linear Regression, Second-Degree Polynomial Regression, Support Vector Machine (SVM), Artificial Neural Network (ANN), Random Forest, and XGBoost. The results demonstrate that while the polynomial model achieves high statistical precision (R² = 0.995, RMSE = 154.62 cSt), the Artificial Neural Network architecture delivers superior robustness (R² = 1.000, RMSE = 154.62 cSt) in capturing complex non-linear thermal behaviors. XGBoost and Random Forest show competitive performance with R² values of 0.90 and 0.91 respectively, while SVM exhibits the highest prediction errors (RMSE ¿ 780 cSt).

Die Inhaltsangabe kann sich auf eine andere Ausgabe dieses Titels beziehen.