基于集成学习算法的页岩油藏压裂效果预测  

Prediction of Fracturing Efficiency in Shale Oil Reservoirs Based on Ensemble Learning Algorithms

作　　者：李丹 

机构地区：[1]重庆理工大学理学院,重庆

出　　处：《传感器技术与应用》2025年第2期117-129,共13页Journal of Sensor Technology and Application

摘　　要：本研究旨在通过集成学习算法对页岩油藏压裂效果进行预测,以提高压裂效果预测的准确性。采用了多种机器学习、深度学习和集成学习模型进行比较,包括线性回归、决策树、XGBoost、LSTM、BP神经网络、投票法、加权投票法、堆叠法等。模型评估指标包括决定系数(R2)、均方误差(MSE)、平均绝对误差(MAE)和均方根误差(RMSE)。实验结果表明,集成学习方法中的堆叠法(Stacking)表现出更优的性能,R2达到0.946,MSE为0.380,MAE为0.383,RMSE为0.616,显著优于单一模型和深度学习模型。通过比较不同模型的性能,本研究表明集成学习方法,特别是堆叠法,能够更有效地提升页岩油藏压裂效果预测的精度。结果为油田开采提供了可靠的预测工具,有助于优化压裂作业和提升生产效率。This study aims to predict the fracturing performance of shale oil reservoirs using ensemble learning algorithms to improve prediction accuracy. Various machine learning, deep learning, and ensemble learning models were compared, including Linear Regression, Decision Tree, XGBoost, LSTM, BP Neural Network, Voting, Weighted Voting, and Stacking. Model evaluation metrics include R2, Mean Squared Error (MSE), Mean Absolute Error (MAE), and Root Mean Squared Error (RMSE). Experimental results indicate that the Stacking method in ensemble learning performs the best, with an R2 of 0.946, MSE of 0.380, MAE of 0.383, and RMSE of 0.616, significantly outperforming both individual models and deep learning models. By comparing the performance of different models, this study demonstrates that ensemble learning methods, especially stacking, can more effectively improve the accuracy of prediction of shale oil reservoir fracturing performance. The results provide reliable predictive tools for oilfield development, helping to optimize fracturing operations and improve production efficiency.

关 键 词：页岩油藏 压裂效果 集成学习 XGBoost 堆叠法 

分 类 号：TE357.1[石油与天然气工程—油气田开发工程]