Uncertainty Quantification

The EVOLVE® workflow tackles a persistent challenge in reservoir management: quantifying the uncertainty in NPV arising from geological and simulation uncertainties, forecast scenarios, and economic parameters. Its innovative approach involves the iterative refinement and reduction of model ensembles as new parameters are introduced across the four stages of the workflow.

The EVOLVE® Workflow

The EVOLVE® workflow follows a four-stage, linear process. Its key innovation lies in the iterative evolution and refinement of model ensembles in a cascading manner as new parameters are introduced at each stage. The resulting final ensemble is both robust and practical for forecasting.

Image
evolve workflow
The EVOLVE workflow is linear. It evolves and reduces model ensembles in a cascading fashion to create a final ensemble that is robust and practical for forecasting.

 

1

Stage 1: Screening Geology

Stage 1 is a geological screening step designed to extract a smaller ensemble of geomodels that preserves the diversity of a much larger set. Rather than evaluating their fit to measured production data, this step focuses on comparing model responses to one another. Model diversity is quantified using K-medoid cluster analysis and visualized through multidimensional scaling (MDS). An efficient flow modeling proxy is essential at this stage to assess a broad and varied ensemble of static geological models, ranging from hundreds to thousands.

Image
evolve stage1
Screening multiple geomodels through 3DSL, followed by clustering, and then extracting a much smaller representative subset of models.

For more information on screening and sensitivity analysis see here

2

Stage 2: Field Level Calibration

The geo-ensemble extracted from Stage 1 is combined with global flow simulation uncertainties (such as OWC depth, relperm functions, PVT properties,...) increasing again the ensemble size.  The goal of this stage is to reduce the size of the ensemble while minimizing the error to historical data and maximizing input parameter diversity. This an optimization problem.

Image
evolve stage2
Include uncertainty in flow simulation parameters such as PVT properties, contact depths, and relperms.  Then optimize on these flow parameters and extract a set of models that are close to history while retaining input parameter diversity.

For more information on screening and sensitivity analysis see here

3

Stage 3: Well-Level Calibration

Stage 3 will yield an ensemble of models that display an acceptable match of the field response. Individual well responses, however, are not guaranteed to exhibit the same match. At this stage, a novel well-level history matching algorithm is used to modify inter-well geology to improve well matches for all models of the ensemble, or a selected subset.

Image
evolve stage3
Automatically select the worst well in each model and then perform well-level history matching on these wells, such that each model in the ensemble has both a good field-level and well-level match.

For more information on screening and sensitivity analysis see here

4

Stage 4: Forecasting & Economics

The final step of the EVOLVE workflow is to use the ensemble of models extracted from stage 2 or stage 3 for forecasting and ecomomic analysis. The ensemble of models is considered robust and practical: robust because the models exhibit diversity, practical because the number of models remains manageable for computing purposes. The ensemble can now be used to investigate short and long term optimization strategies under various economic scenarios.

Image
evolve stage4
For each history matched model in the final ensemble, attach multiple forecast scenarios (base case, infill drilling, polymer injection, etc).  Then for each forecast scenario include uncertainties in NPV by including multiple economic scenarios.