Integrating machine learning-based classification and regression models for solvent regeneration prediction in post-combustion carbon capture: An absorption-based case

• Developed a model for solvent selection and regeneration. • Gradient Boosting achieved 0.942 AUC in predicting solvents. • RF model estimated solvent regeneration with R² of 0.942. • Simulation confirmed RF model's 22.3 MW prediction (22.5 MW). • Supports PCC efficiency for industrial power plant and similar sites. As an industrialized nation with substantial greenhouse gas emissions, Iran faces an urgent need to implement post-combustion carbon capture (PCC) technology. However, conventional PCC systems often struggle with inefficiencies, particularly in solvent selection and regeneration, leading to high energy consumption. This research introduces a primary model comprising two sub-models aimed at optimizing PCC configurations. The first sub-model applies classification techniques Logistic Regression, AdaBoost, Support Vector Classifier, Gradient Boosting, Naive Bayes, Decision Tree, Random Forest, and K-Nearest Neighbors to determine the most suitable solvent based on variables such as pressure, temperature, and concentration. Among these, the Gradient Boosting Classifier demonstrated the best performance, attaining an Area Under the Curve (AUC) score of 0.94. The second sub-model estimates the energy required for solvent regeneration in an industrial-scale power plant using machine learning approaches, including K-Nearest Neighbors (KNN), Random Forest (RF), Decision Tree (DT), Logistic Regression (LR), Bayesian Ridge (BR), and Support Vector Regression (SVR), alongside ensemble strategies like Voting and Stacking. The Random Forest model, optimized via Grid Search Cross-Validation, delivered the most accurate results, achieving an R² of 0.942, a Mean Absolute Error (MAE) of 0.028, and a Mean Squared Error (MSE) of 0.004. The RF model predicted a regeneration energy of 22.3 MW for piperazine, validated by Aspen HYSYS, which gave a value of 22.5 MW. This study shows that using MEA outperforms piperazine as a solvent for PCC.