Mud Gas ML Modeling for Real-Time Formation Pore Pressure
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for estimating pore pressure during drilling operations lack accuracy and reliability, leading to potential drilling incidents such as blowouts and increased non-productive time due to assumptions of linear relationships and reliance on empirical equations.
Innovation Solution
A system utilizing machine learning models, specifically artificial neural networks, to predict pore pressure in real-time based on mud gas data collected during drilling, establishing nonlinear relationships between mud gas and pore pressure data without relying on seismic or surface drilling parameters.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If conventional empirical equations with linear assumptions are used to estimate pore pressure, then the method is simple and easy to implement, but the accuracy of pore pressure estimation deteriorates
Solution Approach 1:
The patent transforms the pore pressure estimation approach by changing from linear empirical equations to a machine learning model that captures non-linear relationships. The system uses mud gas data (gas chromatography data from drilling mud) as input parameters to train a machine learning model, enabling accurate prediction of pore pressure without relying on linear assumptions. This parameter transformation from simple linear variables to complex non-linear patterns resolves the contradiction between simplicity and accuracy.
2Reliability
If accurate pore pressure estimation is achieved using machine learning models with mud gas data, then the accuracy and reliability improve, but the system complexity increases
Solution Approach 1:
The patent replaces traditional mechanical/geological estimation methods with a data-driven machine learning system. Instead of using physical core samples, well logs, or complex geological modeling, the system substitutes these with a computational approach that uses mud gas data processed through machine learning algorithms. This substitution reduces the need for physical infrastructure while increasing computational complexity, resolving the trade-off between reliability and device complexity.
3Reliability
If real-time pore pressure monitoring is implemented during drilling operations, then drilling safety and incident prevention improve, but the time and computational resources required increase
Solution Approach 1:
The patent implements preliminary action by training the machine learning model offline using historical mud gas data and pore pressure measurements before actual drilling operations. The pre-trained model can then rapidly predict pore pressure in real-time during drilling without requiring complex computations at that moment. This preliminary training phase separates the heavy computational work from the real-time operation, resolving the contradiction between safety monitoring and time consumption.
4Measurement precision
If conventional methods using seismic data and surface drilling parameters are used, then comprehensive pore pressure analysis is achieved, but the cost and time for data acquisition increase
Solution Approach 1:
The patent extracts and utilizes only the essential data element - mud gas data from the drilling mud - while eliminating the need for time-consuming seismic surveys, well logs, and surface drilling parameter collections. By focusing on the single most informative parameter (mud gas composition and concentration), the system achieves comprehensive pore pressure analysis without the overhead of multiple data acquisition methods, resolving the contradiction between comprehensiveness and time consumption.
Data Source
AI summary
A system for estimating a pore pressure value associated with a depth of a well subject to drilling operations may include a data repository for storing integrated mud gas and pore pressure data associated with one or more existing wells. The data repository may also store a machine learning (ML) engine. The system may also include one or more hardware processors configured to train a ML model using the ML engine and the integrated mud gas and pore pressure data, to estimate, during the drilling operations, the pore pressure value of a formation zone, at the depth of the well, using the trained ML model and mud gas data associated with a depth value that identifies the depth of the well subject to the drilling operations, and to update a drilling program for a production system based on the estimated pore pressure value.