Reduced-Physics Reservoir Forecasting with RGNet and ML Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing reservoir simulation methods face challenges in efficiently modeling fluid flow dynamics in petroleum reservoirs due to high computational costs, complex model-building processes, and limited data applicability, making it difficult to make timely field decisions.
Innovation Solution
A reduced physics framework using a reservoir graph network (RGNet) model combined with machine learning (ML) algorithms, specifically support-vector-regression with distributed-Gauss-Newton (SVR-DGN), to enhance history-matching and determine well connection patterns, reservoir connectivity, and model parameters, reducing computational cost and improving forecasting accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a full-physics 3D reservoir simulation model is used, then modeling accuracy is improved, but computational cost and model-building time increase significantly
Solution Approach 1:
The patent segments the reservoir system into a network of nodes (wells) and edges (flow paths), reducing the continuous 3D simulation domain into discrete graph elements. This segmentation allows the system to capture essential flow dynamics while dramatically reducing computational complexity and model-building time.
Solution Approach 2:
The patent extracts key flow dynamics from the complex full-physics simulation by identifying and isolating the essential components (well connections, flow paths, pressure gradients). This extraction creates a simplified RGNet model that retains predictive accuracy while eliminating unnecessary computational overhead.
2Measurement precision
If a full-physics 3D reservoir simulation model is used, then modeling accuracy is improved, but device complexity increases
Solution Approach 1:
The patent divides the complex reservoir system into discrete graph elements (nodes and edges), transforming the continuous physics-based model into a simplified network representation. This segmentation reduces model complexity while maintaining accuracy through careful selection of critical flow paths and well connections.
Solution Approach 2:
The patent creates a simplified copy of the reservoir system in the form of a graph network, where the essential dynamics are replicated without the full physical complexity. This copy retains predictive capability while being much easier to build, calibrate, and compute.
3Loss of time
If a reduced-physics model is used, then computational cost is reduced, but physical insights and data applicability are limited
Solution Approach 1:
The patent designs the RGNet model to be universally applicable to various reservoir configurations and data types. The graph network framework can accommodate different well types, flow regimes, and boundary conditions, making the simplified model versatile enough to handle diverse petroleum reservoir problems while maintaining computational efficiency.
Solution Approach 2:
The patent adjusts model parameters (node properties, edge conductances, boundary conditions) to match observed production data and reservoir characteristics. This parameter calibration enables the reduced-physics model to adapt to different reservoir types and data availability scenarios, enhancing its versatility without sacrificing computational speed.
Data Source
AI summary
A method of modeling fluid flow dynamics in a reservoir system, includes receiving observed data for the reservoir system; generating a plurality of model parameters in an initial fluid system model for the reservoir system; performing a plurality of reservoir simulations to determine a well response for the plurality of model parameters; generating a target response using the updated fluid system model for a forecast period; generating a machine learning (ML) model to correct a discrepancy between the target response and the observed data for the forecast period; and determining, using the ML model, a corrected target response for the forecast period.