1. A machine learning-based method for analyzing factors influencing production capacity and production forecasting in fractured tight oil reservoirs.
- Author
-
Tong, Shikai, Wang, Fuyong, Gao, Huanhuan, and Zhu, Weiyao
- Abstract
In recent years, with the deepening of exploration and development theories for tight oil reservoirs and the continuous breakthroughs in key engineering technologies, positive progress has been made in the development of onshore tight oil in China, it has become an important supplement to the growth of petroleum reserves and production. Therefore, identifying the main controlling factors of tight oil reservoir productivity and predicting this productivity are crucial for the development of petroleum resources. The article employs a comprehensive data preprocessing method to address field data, minimizing data errors to the greatest extent possible. In the study of controlling factors and prediction of productivity, various methods such as Pearson correlation coefficient, Random Forest, XGBoost, etc., are compared to enhance the reliability of the results. In addition, root mean square error, mean absolute error, and coefficient of determination are also introduced to evaluate the prediction results of each model, providing an overview of the overall framework of machine learning application in productivity. The study shows that compared with mutual information coefficient, the results of production capacity analysis using Random Forest are more consistent with Pearson correlation coefficient method. The fracturing engineering factors are dominant in the initial production capacity and first-year cumulative production; the initial production capacity is mainly controlled by the soaking time, the stable water cut, the volume of fracturing liquid volume and the formation thickness; the first-year cumulative oil production is mainly dependent on the stable water cut. The XGBoost method yielded the best performance across all three evaluation metrics, indicating that its prediction results are more reliable. In conclusion, machine learning methods can provide significant technical support for the exploration and development of tight oil reservoirs. • Utilize machine learning to discern geological and engineering factors impacting oil production. • Employ Random Forest and Pearson correlation coefficient for production analysis. • Achieve superior accuracy in oil production prediction with XGBoost. [ABSTRACT FROM AUTHOR]
- Published
- 2024
- Full Text
- View/download PDF