河北地质大学学报2025，Vol.48Issue(5)：48-56,9.DOI:10.13937/j.cnki.hbdzdxxb.2025.05.006
TCN-Transformer模型在鄂尔多斯盆地长8储层孔隙度预测精准评价
Accurate Evaluation of Porosity Prediction of Chang 8 Reservoir in Ordos Basin Based on TCN-Transformer Model
刘心如 1曾滨鑫 1刘卫东1
作者信息
- 1. 西安石油大学 地球科学与工程学院,陕西 西安 710065||西安石油大学 陕西省油气成藏地质学重点实验室,陕西 西安 710065
- 折叠
摘要
Abstract
Porosity is a critical parameter in reservoir evaluation.However,traditional prediction methods suffer from limited accuracy due to their inability to effectively model complex nonlinear geological relationships.To improve reservoir parameter prediction,this study proposes a novel hybrid model integrating Temporal Convolutional Network(TCN)and Transformer architectures.Key input logging data were selected based on Pearson correlation coefficients,and genetic algorithm was employed for hyperparameter optimization.The model was applied to the Chang 8 oil layer group in the southwestern Ordos Basin,and its performance was compared with standalone Transformer,CNN,and TCN models.Experimental results demonstrate that the TCN-Transformer model achieves superior performance,exhibiting lower mean absolute error(MAE)and root mean square error(RMSE),along with a coefficient of determination(R2)closer to 1,indicating higher prediction accuracy.Furthermore,the model shows strong generalization capability,yielding the lowest prediction errors on an independent test set without additional training or parameter tuning.This method provides a high-precision tool for porosity prediction,offering practical value for reservoir characterization and development planning.关键词
孔隙度预测/深度学习/TCN-Transformer模型/测井数据/遗传算法Key words
porosity prediction/deep learning/TCN-Transformer model/logging data analysis/genetic algorithm分类
天文与地球科学引用本文复制引用
刘心如,曾滨鑫,刘卫东..TCN-Transformer模型在鄂尔多斯盆地长8储层孔隙度预测精准评价[J].河北地质大学学报,2025,48(5):48-56,9.
