Hydraulic Fracture Half-Length History Matching with Transfer Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Determining fracture half-length in subsurface regions is challenging due to limited data availability and the complexity of hydraulic fracturing in unconventional shales, leading to uncertainties in production estimates and the impact of nearby wells, which complicates the determination of optimal well locations and fracturing strategies.
Innovation Solution
A data processing system utilizing transfer learning and machine learning models, combined with hydraulic fracturing simulation and probabilistic sampling, to predict fracture half-lengths based on subsurface properties, enabling accurate production forecasting and optimal well placement.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional history matching methods are used to determine fracture half-length, then measurement precision can be achieved with sufficient data, but the process becomes computationally intensive and time-consuming due to multiple simulation iterations
Solution Approach 1:
The patent performs preliminary actions by pre-computing sensitivity matrices and preparing training data from simulation results before actual history matching is needed. The machine learning model is trained in advance on simulated fracture half-length data, enabling rapid predictions during actual well evaluation without repeating time-consuming simulation iterations.
Solution Approach 2:
The patent replaces the mechanical simulation-based history matching process with a machine learning-based prediction system. Instead of iteratively running physical reservoir simulation models to match production history, the system uses trained ML models that instantly predict fracture half-lengths based on input parameters, substituting computational mechanics with data-driven algorithms.
2Reliability
If multiple hydraulic fracturing simulations are run to account for subsurface uncertainty, then reliability of production estimates improves, but device complexity and computational resources increase significantly
Solution Approach 1:
The patent creates copies of the complex reservoir simulation system in the form of trained machine learning models. These ML models capture the essential relationships between subsurface properties, fracturing parameters, and production outcomes without requiring the full complexity of the original simulation system to be executed repeatedly, thus maintaining reliability while reducing operational complexity.
Solution Approach 2:
The patent changes the fundamental parameters of the system by transforming physical simulation outputs into machine learning model inputs. Instead of running multiple complex simulations with varying subsurface realizations, the system uses probabilistic parameter sampling combined with trained ML models to efficiently explore uncertainty and generate reliable production estimates.
3Productivity
If fracture half-length is increased to maximize hydrocarbon recovery, then production improves, but the complexity of determining optimal fracture length increases due to limited subsurface data
Solution Approach 1:
The patent introduces machine learning models as intermediary systems between subsurface property inputs and fracture half-length predictions. These intermediaries process limited and uncertain subsurface data through trained algorithms, transforming vague geological inputs into specific, actionable fracture design recommendations without requiring direct complex measurements.
Solution Approach 2:
The patent implements feedback mechanisms where production data from actual wells is used to retrain and refine the machine learning models. This continuous feedback loop allows the system to learn from real-world performance and improve its predictions, making the determination of optimal fracture lengths more accurate despite limited initial subsurface data.
Data Source
AI summary
Methods and systems are configured for obtain reservoir data representing one or more features of a reservoir. The reservoir data include a set of values that satisfy a probability distribution associated with that feature. The process includes performing a geological simulation of hydraulic fracturing in the reservoir, the geological simulation generating a production estimate for a well based on the reservoir data, the production estimate associated with the one or more features; generating training data using the production estimate associated with the one or more features. The process includes training, using the training data, a machine learning model to predict a fracture half-length value of the reservoir, the fracture half-length value corresponding to the hydraulic fracturing represented by the training data. The process includes determining, based on the training, a history-matched value for the fracture half-length of the reservoir.