Machine Learning Surrogate for Multilateral Well ICV Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for predicting multilateral well performance are inefficient and unable to accurately optimize Inflow Control Valve (ICV) settings, leading to suboptimal hydrocarbon production from 'tight' reservoirs.
Innovation Solution
A method using machine learning with physics-based models to predict multilateral well performance by training a neural network with simulation data from physics-based models, incorporating well completion, inflow control valve, and reservoir attributes to optimize ICV settings and predict production parameters.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If physics-based models are used to predict multilateral well performance, then prediction accuracy is improved, but computation time and storage requirements increase significantly
Solution Approach 1:
The patent pre-generates comprehensive simulation data covering the full range of possible operating conditions and reservoir parameters before actual prediction is needed. This training dataset captures the complex physics-based model outputs in advance, allowing the machine learning model to learn from this pre-computed information and make rapid predictions without re-running the computationally intensive physics-based models during operational decision-making
Solution Approach 2:
The patent creates a machine learning model that replicates the behavior of the physics-based model by training it on simulation data generated from the physics-based model. The trained ML model serves as a simplified copy or surrogate that mimics the complex physics-based model's predictions but executes much faster, enabling real-time optimization without the computational burden of the original physics-based model
2Measurement precision
If physics-based models are used to predict multilateral well performance, then prediction accuracy is improved, but storage requirements increase by 99%
Solution Approach 1:
The patent replaces the need to store large volumes of raw simulation data and physics-based model outputs with a compact machine learning model. The trained ML model encapsulates the essential patterns and relationships learned from the simulation data in a compressed form, requiring minimal storage while maintaining prediction accuracy
Solution Approach 2:
The patent transforms the complex, high-dimensional simulation data into a streamlined machine learning model with optimized parameters. By changing the representation from raw simulation outputs to trained model parameters, the storage requirements are dramatically reduced while preserving the predictive capabilities
3Productivity
If traditional methods are used for ICV optimization, then implementation is simpler, but hydrocarbon production efficiency remains suboptimal
Solution Approach 1:
The patent implements an iterative optimization process where the machine learning model predicts production outcomes for different ICV configurations, and these predictions feed back into adjusting the ICV settings to maximize hydrocarbon recovery. This closed-loop feedback system continuously refines the optimization based on predicted performance, enabling superior productivity compared to static traditional methods
Solution Approach 2:
The patent transitions from static, fixed ICV settings to dynamic, adaptive optimization using machine learning. The system can adjust ICV configurations in response to changing reservoir conditions and production targets, enabling real-time optimization that adapts to dynamic operational requirements and maximizes productivity throughout the well's lifecycle
Data Source
AI summary
A system and method for machine learning with physics-based models to predict multilateral well performance are provided. An exemplary method enables obtaining data associated with well completion, data associated with inflow control valves, and reservoir attributes of multilateral wells. Production scenarios are generated based on the data associated with well completion, the data associated with inflow control valves, and the reservoir attributes of the multilateral wells. The production scenarios are input into a physics-based model of the multilateral wells, and simulation data associated with the multilateral wells output from the physics-based model is obtained. A neural network based machine learning model is trained using the simulation data associated with the multilateral wells and target parameters associated with the multilateral wells, wherein the trained machine learning model is configured to predict multilateral well production parameters.