A method and system for multi-well joint prediction of water cut of oil wells in an extra-high water cut stage

By using a cascaded TCN-LSTM structure, the complex problem of joint prediction of multiple wells in oil wells during ultra-high water cut is solved. This enables joint modeling of multiple wells in oil well scenarios during ultra-high water cut, adapts to complex dynamics at the well level, and improves prediction performance and consistency.

CN122242786APending Publication Date: 2026-06-19CHINA UNIV OF PETROLEUM (EAST CHINA)
View PDF 4 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
CHINA UNIV OF PETROLEUM (EAST CHINA)
Filing Date
2026-05-25
Publication Date
2026-06-19

AI Technical Summary

Technical Problem

Existing technologies struggle to achieve multi-well joint modeling in oil well scenarios with extremely high water cut, failing to effectively characterize complex well-level dynamics. Single-well time-series models cannot utilize shared patterns across multiple wells, and multi-well deep learning solutions often prioritize production rather than water cut.

Method used

A cascaded TCN-LSTM structure is adopted. By acquiring historical production dynamic data from multiple oil wells, data preprocessing and normalization are performed to construct supervised learning samples. These samples are then merged to form a joint training set. The TCN module is used to extract local fluctuation features and stage change features, while the LSTM module characterizes long-term dependencies, enabling joint prediction of multiple wells.

Benefits of technology

While maintaining the time sequence of individual wells, the model can be trained using historical dynamic data from multiple oil wells in a unified manner, adapting to the non-stationary and nonlinear changes in water cut of individual wells during the ultra-high water cut period, thereby improving the consistency of the model in depicting the overall trend and its engineering applicability.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122242786A_ABST
    Figure CN122242786A_ABST
Patent Text Reader

Abstract

This invention discloses a multi-well joint prediction method and system for water cut in oil wells during ultra-high water cut stages, relating to the field of oil and gas field development. The method includes the following steps: S1: Acquire historical production dynamic data from multiple wells in the ultra-high water cut development stage; S2: Sort the historical production dynamic data by date; S3: Perform sliding sampling on the standardized time-series data of each oil well using a preset historical time window; S4: Divide the data of each single well into training and test sets according to time sequence; S5: Merge the data from multiple single wells to form a joint training set and a joint test set; S6: Construct a tandem TCN-LSTM prediction model and train the model; S7: Use the trained model to infer the predicted water cut value for each oil well at the next time step. This invention addresses the complex fluctuation characteristics of water cut in oil wells during ultra-high water cut stages by predicting the water cut at the next time step based on well-level historical production dynamic data, making it more suitable for handling the non-stationary and nonlinear changes in water cut of single wells during ultra-high water cut stages.
Need to check novelty before this filing date? Find Prior Art