Scenario Planning
One of the most powerful applications of digital twins is scenario planning - the ability to ask "what if?" questions and see the predicted outcomes before committing resources. Instead of relying on intuition or spreadsheet approximations, engineers can run rigorous simulations of alternative strategies in hours rather than months.
Common Scenario Types
Development Scenarios
Compare different field development plans: number of wells, well types (horizontal vs vertical), drilling sequence, and facility sizing to find the highest NPV strategy.
Example: Plan A (20 horizontal wells) yields 180 MMbbl recovery at $2.1B CAPEX. Plan B (30 vertical wells) yields 165 MMbbl at $1.6B CAPEX. The twin helps quantify the trade-off.
Operational Scenarios
Evaluate real-time decisions: What happens if we shut Injector-W03 for workover? How does increasing gas lift on Well-B07 affect total field production? What is the impact of a 2-week compressor shutdown?
Economic Scenarios
Test sensitivity to oil price, exchange rates, and cost assumptions. At what oil price does the infill drilling campaign break even? Should we accelerate development in a high-price environment?
Risk & Uncertainty Scenarios
Run Monte Carlo simulations varying uncertain parameters (porosity, permeability, aquifer size) to generate P10/P50/P90 production forecasts and quantify subsurface risk.
Rapid Scenario Evaluation with Proxy Models
Full reservoir simulations can take hours or days to run, limiting the number of scenarios that can be evaluated. Proxy models (also called surrogate models or response surfaces) are ML models trained on simulation outputs that approximate the simulator's behaviour in seconds.
Full Simulation
Hours per run → 50 scenarios evaluated in a week
Proxy Model
Seconds per run → 10,000 scenarios evaluated in an hour
Use case: A field development team needs to select the optimal location for 8 infill wells from 200 candidate locations. Running all possible combinations in the full simulator would take years. Instead, they train a proxy model on 5,000 representative simulation runs and screen all 200 locations in 2 hours, identifying the top 15 candidates. These 15 are then verified with full simulation runs - a process that takes just 3 days.