Fast front tracking in EOR flooding simulation on coarse grids
AssigneeSchlumberger Technology Corporation
InventorsBarys Andreevich Samson, Sophie Nazik Godefroy, Soham Sheth, Christopher Marooney
Life of this patent
Estimating the term…
Abstract
The present disclosure provides a workflow for modelling EOR flooding operations performed on a reservoir by separating front tracking from the reservoir simulation process, so that the front's position and topology evolves in parallel with the coarse grid simulation, through modifications using machine-learning-trained correlations.
Claims27
A method of modeling or simulating enhanced-oil recovery (EOR) flooding operations performed on a reservoir, comprising:
- collecting training data based on simulation of EOR flooding operations using a first grid having a relatively high resolution;
- using the training data to configure and train a machine learning system;
- performing simulation of EOR flooding operations using a second grid having a relatively low resolution than the relatively high resolution of the first grid;
- and
- using output data generated by the simulation using the second grid as input to the machine learning system to predict at least one parameter characterizing a displacing agent front in the second grid for a time step in the simulation using the second grid, wherein the at least one parameter characterizing the displacing agent front comprises a front speed for the time step and the front speed is used during execution of the simulation using the second grid to adjust reservoir simulation parameters associated with at least one cell of the second grid, wherein the reservoir simulation parameters are used to compute a subsequent time step of the simulation using the second grid.
A system for modeling enhanced-oil recovery (EOR) flooding operations performed on a reservoir, comprising:
- at least one processor configured to:
- perform simulation of enhanced-oil recovery (EOR) flooding operations using a first grid having a relatively low resolution;
- and
- use output data generated by the simulation as input to a machine learning system to predict at least one parameter characterizing a displacing agent front in the first grid for a time step in the simulation, wherein the at least one parameter characterizing the displacing agent front comprises a front speed for the time step and the front speed is used during execution of the simulation to adjust reservoir simulation parameters associated with at least one cell of the first grid, wherein the reservoir simulation parameters are used to compute a subsequent time step of the simulation,
- wherein the machine learning system is trained on training data collected from simulation of EOR flooding operations using a second grid having a relatively high resolution.
A method of modeling or simulating enhanced-oil recovery (EOR) flooding operations performed on a reservoir, comprising:
- collecting training data based on simulation of EOR flooding operations using a first grid having a relatively high resolution;
- using the training data to configure and train a machine learning system;
- performing simulation of EOR flooding operations using a second grid having a relatively low resolution than the relatively high resolution of the first grid, wherein a resolution of the second grid is maintained during execution of the simulation using the second grid;
- and
- using output data generated by the simulation using the second grid as input to the machine learning system to predict at least one parameter characterizing a displacing agent front in the second grid for a time step in the simulation using the second grid, wherein the at least one parameter characterizing the displacing agent front comprises a front speed for the time step and the front speed is used during execution of the simulation using the second grid to adjust reservoir simulation parameters associated with at least one cell of the second grid, wherein the reservoir simulation parameters are used to compute a subsequent time step of the simulation using the second grid.
Dependencies are inferred from the claim text.