A method for predicting formation pressure based on an RTH improved CNN-LSTM model

CN122153780APending Publication Date: 2026-06-05LIAONING UNIVERSITY OF PETROLEUM AND CHEMICAL TECHNOLOGY

Patent Information

Authority / Receiving Office: CN · China
Patent Type: Applications(China)
Current Assignee / Owner: LIAONING UNIVERSITY OF PETROLEUM AND CHEMICAL TECHNOLOGY
Filing Date: 2026-02-12
Publication Date: 2026-06-05

AI Technical Summary

Technical Problem

During drilling, existing technologies struggle to accurately predict formation pore pressure, leading to complex downhole failures such as wellbore instability, drilling fluid loss, blowouts, and well kicks, increasing construction costs and risks. These problems are particularly pronounced in deep and ultra-deep wells.

Method used

An improved CNN-LSTM model is adopted, which combines dynamic drilling data and static logging data. The model weights are optimized by the Red-tailed Eagle optimization algorithm to extract local geological features and vertical time series features, thereby improving the accuracy of formation pressure prediction.

Benefits of technology

Under small sample data conditions, it significantly improved the prediction accuracy of abnormal formation pressure, reduced drilling risks and costs, optimized wellbore structure, and improved mechanical drilling rate and well control safety.

✦ Generated by Eureka AI based on patent content.

Smart Images

Figure CN122153780A_ABST

Patent Text Reader

Abstract

The application provides a method for stratum pressure prediction based on an RTH improved CNN-LSTM model, comprising the following steps: obtaining dynamic drilling data and static logging data and fusing the same as inputs of the CNN-LSTM model, so as to extract local geological features through a CNN model and extract longitudinal time sequence features through an LSTM model; and optimizing the weights of the CNN model and the LSTM model by using a red-tailed hawk optimization algorithm, and simultaneously determining the optimal convolution kernel of the CNN model and the optimal number of neurons of the LSTM model, so as to realize the adjustment of the local geological features and the longitudinal time sequence features and obtain fused features for pressure prediction. According to the method, the prediction accuracy of stratum abnormal pressure can be improved through the improved CNN-LSTM fusion model on the basis of small sample drilling and logging data.

Need to check novelty before this filing date? Find Prior Art