A Transformer-based method and system for predicting pore pressure

By using a Transformer-based method for pore pressure prediction, and employing a multi-head cross-attention mechanism and sinusoidal position coding, the problem of inaccurate pore pressure prediction is solved, achieving high-precision and efficient pore pressure prediction that is adaptable to complex geological conditions.

CN123069908APending Publication Date: 2026-11-10CHINA UNIV OF MINING & TECH (BEIJING)
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611370032.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-09-04
Publication Date
2026-11-10

AI Technical Summary

Technical Problem

Existing technologies suffer from inaccurate predictions in pore pressure forecasting, particularly in handling complex nonlinear couplings and global dependencies in well logging data.

Method used

A Transformer-based approach is adopted, which uses a multi-head cross-attention mechanism for feature fusion, and combines sinusoidal position encoding and a pre-trained Transformer model to capture the complex nonlinear and global dependencies between well logging data for pore pressure prediction.

Benefits of technology

It improves the accuracy and efficiency of pore pressure prediction, can adapt to complex geological scenarios, reduces the reliance on artificial feature engineering, and enhances the generalization performance of the model.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123069908A_ABST
    Figure CN123069908A_ABST
Patent Text Reader

Abstract

This application discloses a Transformer-based pore pressure prediction method and system. The method includes: acquiring multiple pre-collected well logging data; fusing features of the multiple well logging data through a multi-head cross-attention mechanism to obtain fused features; expanding the dimensions of the fused features to obtain an expanded feature vector; encoding the expanded feature vector based on the position index and dimension index in the expanded feature vector using a sinusoidal position encoding formula to obtain an encoded feature vector; extracting features from the encoded feature vector based on a pre-trained Transformer model to obtain a target feature vector; the target feature vector is used to characterize the complex nonlinear and global dependencies between multiple well logging data; and predicting the target feature vector using the Transformer model to obtain a predicted pore pressure value.
Need to check novelty before this filing date? Find Prior Art