An Improved XGBoost Model for Development Parameter Optimization and Production Forecasting in CO₂ Water-Alternating-Gas Processes: A Case Study of Low Permeability Reservoirs in China

The pronounced heterogeneity and geological complexity of low-permeability reservoirs pose significant challenges to parameter optimization and performance prediction during the development of CO₂ water-alternating-gas (CO₂-WAG) injection processes. This study introduces a predictive model based on the Extreme Gradient Boosting (XGBoost) algorithm, trained on 1225 multivariable numerical simulation cases of CO₂-WAG injection. To enhance the model’s performance, four advanced metaheuristic algorithms—Collective Parallel Optimization (CPO), Grey Wolf Optimization (GWO), Artificial Hummingbird Algorithm (AHA), and Black Kite Algorithm (BKA)—were applied for hyperparameter tuning. Among these, the CPO algorithm demonstrated superior performance due to its ability to balance global exploration with local exploitation in high-dimensional, complex optimization problems. Additionally, the integration of Chebyshev chaotic mapping and Elite Opposition-Based Learning (EOBL) strategies further improved the algorithm’s efficiency and adaptability, leading to the development of the ICPO (Improved Crowned Porcupine Optimization)-XGBoost model. Rigorous evaluation of the model, including comparative analyses, cross-validation, and real-case simulations, demonstrated its exceptional predictive capacity, with a coefficient of determination of 0.9894, a root mean square error of 2.894, and errors consistently within ±2%. These results highlight the model’s robustness, reliability, and strong generalization capabilities, surpassing traditional machine learning methods and other state-of-the-art boosting-based ensemble algorithms. In conclusion, the ICPO-XGBoost model represents an efficient and reliable tool for optimizing the CO₂-WAG process in low-permeability reservoirs. Its exceptional predictive accuracy, robustness, and generalization capability make it a highly valuable asset for practical reservoir management and strategic decision-making in the oil and gas industry.