Subspace Ensemble Kalman Filter for Non-Gaussian Reservoir Simulation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The Ensemble Kalman Filter (EnKF) faces challenges such as ensemble collapse and the inability to honor geostatistical properties of non-Gaussian random fields, particularly in large-scale simulation models of subterranean reservoirs, leading to poor uncertainty quantification and inefficient computational burden when dealing with multiple geological scenarios.
Innovation Solution
The subspace Ensemble Kalman Filter constrains each ensemble member to a specific subspace using different parameterizations, ensuring that geostatistical properties are retained throughout updates, and employs kernel principal component analysis or Karhunen-Loeve expansion for efficient handling of non-Gaussian fields, mitigating ensemble collapse and improving uncertainty quantification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If standard Ensemble Kalman Filter is applied to large-scale reservoir models, then ease of implementation and efficient uncertainty quantification are achieved, but ensemble collapse occurs and geostatistical properties are lost
Solution Approach 1:
The patent segments the reservoir model into multiple geological scenarios or realizations, each representing different geological interpretations. Instead of using a single ensemble that collapses to one solution, multiple ensembles are maintained separately, each preserving its unique geostatistical properties and geological scenario. This segmentation prevents the ensemble collapse problem by ensuring diverse geological interpretations are maintained throughout the data assimilation process.
Solution Approach 2:
The patent applies different geostatistical parameters and properties to different ensemble members, allowing each to maintain local geological characteristics. This local quality approach ensures that each ensemble member preserves its specific geostatistical properties (such as permeability distributions, porosity characteristics) while still being updated with production data, preventing the homogenization that leads to ensemble collapse.
2Reliability
If multiple EnKF are applied for each geological scenario, then geostatistical properties are preserved, but computational burden increases N times over standard EnKF
Solution Approach 1:
The patent merges multiple geological scenarios into a unified framework where multiple ensembles are updated simultaneously using the same production data. This combining approach allows all geological scenarios to be processed together in a single computational framework, achieving the same results as running separate EnKF for each scenario but with improved computational efficiency through shared calculations and data structures.
Solution Approach 2:
The patent creates a universal EnKF framework that can handle multiple geological scenarios simultaneously. This multi-functional system processes diverse geological interpretations within a single algorithmic structure, allowing the same code and computational resources to serve multiple geological scenarios, thereby reducing the N-fold computational burden while preserving geostatistical properties across all scenarios.
3Measurement precision
If Ensemble Kalman Filter updates are applied repeatedly, then data matching improves, but all ensemble members converge towards the same member reducing variability
Solution Approach 1:
The patent applies preliminary geostatistical conditioning to each ensemble member before the EnKF update process begins. By pre-establishing distinct geostatistical properties and geological scenarios for each ensemble member, the method creates a foundation that resists convergence during repeated updates. This preliminary action ensures that even as ensemble members are updated multiple times to improve data matching, they retain their unique geological characteristics and variability.
Data Source
AI summary
A computer-implemented method, system, and computer program product are disclosed for updating simulation models of a subterranean reservoir. An ensemble of reservoir models representing a subterranean reservoir having non-Gaussian characteristics is provided and the ensemble of reservoir models is updated using a subspace ensemble Kalman filter. Kemal principle component analysis parameterization or K-L expansion parameterization can be used to update the ensemble of reservoir models.