Adjoint-Based Conditioning of Geologic Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Process-based models for meandering channel systems in hydrocarbon exploration face challenges in conditioning due to discontinuities and scarcity of field data, leading to instability and inaccurate sensitivity information, particularly during cut-off events.
Innovation Solution
The use of an adjoint-based gradient optimization method with hierarchical conditioning to stabilize and enhance the accuracy of meandering channel simulations by redistributing points on the channel centerline and modifying the objective function to handle discontinuities, allowing for the computation of sensitivity information and improving the matching of predicted data with known field data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If process-based models are used to simulate meandering channel systems, then geologically realistic models can be generated by time-advancing governing equations, but the conditioning becomes unstable and sensitivity information becomes inaccurate due to discontinuities and scarcity of field data
Solution Approach 1:
The patent applies preliminary action by redistributing points on the channel centerline before performing the adjoint-based sensitivity analysis. This preprocessing step ensures that the computational mesh is properly positioned to handle discontinuities such as cut-off events, thereby preventing instability in the adjoint model and ensuring accurate sensitivity information can be computed despite the scarcity of field data
Solution Approach 2:
The patent introduces an intermediary objective function that specifically accounts for discontinuities in the meandering channel system. This modified objective function acts as a mediator between the process-based model and the adjoint-based optimization, allowing the sensitivity analysis to proceed accurately even in the presence of discontinuities and limited field data
2Productivity
If gradient-based optimization methods are used to solve inverse problems, then sensitivity information is required to locate local extrema, but discontinuities in the model cause instability and inaccurate sensitivity computations
Solution Approach 1:
The patent performs preliminary redistribution of channel centerline points before executing the gradient-based optimization. This ensures that the computational framework is prepared to handle discontinuities, allowing the sensitivity computations to proceed stably and accurately throughout the optimization process for locating local extrema
Solution Approach 2:
The patent introduces a modified objective function as an intermediary that specifically addresses discontinuities. This intermediary function enables gradient-based optimization to compute sensitivity information reliably even when the underlying meandering channel model exhibits discontinuous behavior during cut-off events
3Device complexity
If field data is sparsely available to constrain process-based models, then model complexity can be reduced, but the ability to condition the model accurately deteriorates
Solution Approach 1:
The patent introduces a specially designed objective function as an intermediary that compensates for sparse field data. This function incorporates physical constraints and discontinuity handling mechanisms that allow accurate conditioning to be achieved even with limited data, avoiding the need for overly complex conditioning procedures
4Reliability
If cut-off events are modeled in meandering channel systems, then geological realism is improved, but discontinuities arise that cause instability in adjoint-based sensitivity analysis
Solution Approach 1:
The patent applies preliminary redistribution of computational points along the channel centerline before performing adjoint analysis. This preparation ensures that the computational framework is ready to handle the discontinuities introduced by cut-off events, maintaining stability in the adjoint solution while preserving geological realism
Solution Approach 2:
The patent introduces a modified objective function as an intermediary that specifically addresses the discontinuities caused by cut-off events. This intermediary enables the adjoint-based sensitivity analysis to proceed stably while maintaining the geological realism of the meandering channel model
Data Source
Figure 1
Figure 2~3
Figure 4~5
AI summary
A method for correlating data predicted by a process- or physics-based geologic model to describe a subsurface region with obtained data describing the subsurface region. Data is obtained describing an initial state of the subsurface region. Data describing a subsequent state of the subsurface region is predicted. The predicted data is compared with the obtained data taking into account whether the obtained data or the predicted data represent a discontinuous event. A sensitivity of the predicted data is determined if the predicted data is not within an acceptable range of the obtained data. The data describing the initial state of the subsurface region is adjusted based on the sensitivity before performing a subsequent iteration of predicting data describing the subsequent state of the subsurface region. A representation of the subsurface region based on the data describing the subsequent state of the subsurface region is outputted.