In-situ Porous Medium Property Prediction via Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current core flooding analysis techniques are time-consuming and require human intervention, struggling to accurately characterize relative permeability and capillary pressure due to core heterogeneity, capillary effects, and gravity forces, leading to inefficiencies in oil recovery predictions.
Innovation Solution
An automated and accelerated post-processing method using real-time core flooding data and machine learning prediction models executed on an AI-based processor for in-situ estimation and dynamic visualization of porous medium properties, specifically relative permeability and capillary pressure, significantly reducing analysis time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If traditional post-processing analysis methods are used to obtain relative permeability and capillary pressure data, then measurement precision is maintained, but analysis time is excessively long and human intervention is required
Solution Approach 1:
The patent replaces traditional mechanical post-processing analysis methods with an automated electronic system that uses machine learning models to predict relative permeability and capillary pressure curves in real-time during core flooding experiments, eliminating manual intervention and dramatically reducing analysis time
Solution Approach 2:
The system enables self-service by automatically processing core flooding data through embedded machine learning models without requiring human operators to perform time-consuming post-processing analysis, allowing the system to generate accurate SCAL data independently and continuously
2Measurement precision
If commercial simulation models are used to estimate capillary pressure and relative permeability, then measurement precision is improved, but device complexity and human intervention requirements increase
Solution Approach 1:
The patent segments the complex simulation problem into manageable components by using pre-trained machine learning models that have been developed offline, allowing the online system to simply execute predictions without dealing with the full complexity of physical simulation models
Solution Approach 2:
The system creates a simplified copy of the complex physical simulation process through machine learning models that replicate the behavior of commercial simulation software, enabling accurate predictions without requiring the full computational infrastructure and human expertise of the original simulation systems
3Reliability
If steady state core flooding experiments are conducted, then reliable SCAL data is obtained, but the effect of capillarity and gravity forces leads to wrongful estimation of parameters
Solution Approach 1:
The system continuously monitors core flooding data in real-time and uses machine learning models to predict relative permeability and capillary pressure curves, providing immediate feedback that allows for dynamic adjustment and interpretation of experimental results, accounting for capillary and gravity effects throughout the experiment rather than only in post-processing
Solution Approach 2:
The machine learning models are pre-trained on extensive datasets that include various capillary and gravity effects, enabling the system to automatically compensate for these phenomena during real-time analysis without requiring separate correction steps or human expertise in interpreting their effects
Data Source
AI summary
There is provided a method and apparatus for accelerated in-situ prediction and dynamic visualization of characteristics of a porous medium, including an input source for inputting data from computer-aided simulations and real-time core flooding experiments, an embedded hardware unit comprising of a processor running a prediction model with inferences of porous medium samples, and a human-machine interface comprising a display unit for displaying the estimated plurality of characteristics of the porous medium and an input unit for accepting commands from a user. The input source is in real-time communication with the embedded hardware unit and the display unit and the apparatus reduces a total analysis time taken for characterizing the porous medium. Further, the porous medium is a rock sample and the plurality of properties of the porous medium comprises the relative permeability (Kr) and capillary pressure (Pc) characteristics of the porous medium.