Physics of Fluids, Journal Year: 2025, Volume and Issue: 37(4)
Published: April 1, 2025
The complex pore structure and weak logging responses in low-porosity, low-permeability reservoirs pose challenges for traditional methods to identify reservoir fluids. Additionally, the imbalance fluid categories leads classification bias toward majority classes, reducing recognition performance minority categories. To address these challenges, this paper proposes a Reservoir Fluid Identification method that combines an improved synthetic over-sampling technique (SMOTE) with revisiting mobile convolutional neural network (CNN) from vision transformer (ViT) perspective (RepViT)-transformer hybrid model. Logging data is processed using sliding window generate sequence samples preserve stratigraphic information while optimizing sample CNN. SMOTE algorithm employs symmetric generation strategy expand Minority Class Samples maintaining distribution characteristics, mitigating class problem. model integrates local feature extraction of RepViT global dependency capturing capabilities transformer, enhancing Classification Accuracy Stability. Applied Dagang Oilfield, outperformed bidirectional long short-term memory (BiLSTM) models both overall identification, achieving 94.41% accuracy on test set—4.92% 10.24% higher than BiLSTM respectively. These results demonstrate method's effectiveness identifying fluids reservoirs, offering novel approach evaluation conventional data.
Language: Английский