物探化探计算技术2025，Vol.47Issue(6)：923-930,8.DOI:10.12474/wthtjs.20240923-0003
基于CAM-ResNet-XGBoost模型的产能预测评价方法研究
Research on logging productivity prediction method based on CAM-ResNet-XGBoost model
袁秋霞 1沈东义 1郭林 1崔荣升1
作者信息
- 1. 中国海洋石油有限天津分公司,天津 300459
- 折叠
摘要
Abstract
Production capacity prediction is a comprehensive evaluation technique for the oil production capabilities of reservoirs.It is critically important for the exploration and development of oil and gas fields.A production capacity prediction model based on channel attention mechanism(CAM),deep residual neural network(ResNet),and XGBoost is proposed to address the low accuracy of production capacity prediction and evaluation caused by the dual influence of lithology and physical properties on mid to deep low porosity and low-permeability reservoirs.Thirteen parameters with high correlation to oil production are selected,and a production capacity prediction model is constructed and trained using exploration well data from an oilfield in Bohai to predict the oil production of new wells.The research results indicate that the proposed CAM-ResNet-XGBoost model can effectively extract valid features from diverse and highly overlapping parameter information.Compared with models such as BP neural network,AlexNet,VGGNet,and ResNet,the CAM-ResNet-XGBoost model has higher prediction accuracy,with an optimal prediction accuracy of 79.6%,which is 8.6%higher than the benchmark model.关键词
测井资料/产能预测/自注意力/残差网络/特征提取Key words
logging data/production capacity prediction/self attention/residual network/feature extraction分类
能源科技引用本文复制引用
袁秋霞,沈东义,郭林,崔荣升..基于CAM-ResNet-XGBoost模型的产能预测评价方法研究[J].物探化探计算技术,2025,47(6):923-930,8.
