Reservoir Proxy Modeling for Faster Oilfield Parameter Tuning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current oilfield modeling and simulation techniques are slow and costly, leading to delays in updating models and making accurate determinations for drilling and production operations, often relying on rough estimations due to the high complexity and computational resource requirements of subsurface volume modeling.
Innovation Solution
The development of a system and method that trains a proxy model to predict outputs from a reservoir model, allowing for real-time or on-time updates and predictions of performance indicators for oilfield operations, using machine learning and AI-driven enhancements to facilitate agile reservoir modeling and data processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If complex reservoir models are built to accurately depict subsurface geology and fluid migration, then model accuracy is improved, but computing time and resources increase significantly
Solution Approach 1:
The patent creates a proxy model that copies the essential input-output relationships of the complex reservoir model. The proxy model is trained on a subset of reservoir model inputs and outputs, then used to rapidly predict outputs for new inputs without running the full complex model, thus achieving fast predictions with acceptable accuracy.
Solution Approach 2:
The proxy model is trained in advance using a training set generated from the complex reservoir model. This preliminary training phase captures the relationship between inputs and outputs, so that during actual operation, predictions can be made rapidly without repeating the full complex simulation.
2Measurement precision
If complex reservoir models are built to accurately depict subsurface geology, then model accuracy is improved, but computational resources and cost increase significantly
Solution Approach 1:
The proxy model serves as a simplified copy that replicates the essential functionality of the complex reservoir model. It uses machine learning algorithms with far fewer computational resources to approximate the input-output relationships, dramatically reducing energy and resource consumption while maintaining acceptable prediction accuracy.
Solution Approach 2:
The proxy model acts as a computationally inexpensive alternative to the expensive complex reservoir model. While the proxy model may have lower precision in some edge cases, it provides sufficiently accurate predictions for operational decision-making at a fraction of the computational cost.
3Productivity
If rough estimations are used for operating parameters to reduce computational burden, then processing speed is improved, but prediction accuracy deteriorates
Solution Approach 1:
Rather than using rough estimations or simple heuristics, the patent creates a proxy model that copies the predictive capabilities of the complex reservoir model. The machine learning-based proxy model learns the complex relationships between inputs and outputs, providing accurate predictions at high processing speeds suitable for real-time or near-real-time operational decision-making.
Solution Approach 2:
The patent transforms the complex reservoir modeling problem into a machine learning prediction problem by changing the approach from physics-based simulation to data-driven prediction. This parameter change in the modeling methodology enables both high speed and good accuracy by leveraging patterns learned from training data.
Data Source
AI summary
A method includes training a proxy model to predict output from a reservoir model of a subterranean volume, receiving data representing an oilfield operation performed at least partially in the subterranean volume, predicting one or more performance indicators for the oilfield operation using the proxy model, and updating the reservoir model based at least in part on the one or more performance indicators predicted in the proxy model.