Transfer Learning for Oil Reservoir History Matching Initialization

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Conventional history matching methods for oil reservoirs face inefficiencies and inaccuracies, particularly in large-scale simulations, due to high computational workload and the limitations of gradient-based and gradient-free methods, and the need for extensive parameter adjustments, which hinder widespread application and precision.

Innovation Solution

An automatic history matching system utilizing transfer learning to reinitialize the population based on historical data, integrating with evolutionary optimization algorithms, to construct an initial population efficiently and accurately, reducing computational complexity and improving precision.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If conventional history matching methods adjust massive parameters manually with tedious and meticulous operations, then the model can be corrected to match true reservoir conditions, but the workload is extremely great and the efficiency is extremely low

Engineering Contradiction:

Improvematching precisionVSAvoidworkload

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The system implements automatic history matching where the algorithm autonomously adjusts reservoir parameters without manual intervention. The automatic history matching module automatically modifies static parameters based on optimization algorithms, eliminating the need for manual parameter adjustment while maintaining matching precision.

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The patent replaces manual mechanical adjustment operations with automated computational algorithms. The system uses optimization algorithms (gradient methods, gradient-free methods, or hybrid approaches) to automatically adjust parameters, substituting human manual operations with computational automation.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Reliability

If the Newton method stores and calculates a Hessian matrix to solve optimization problems, then the optimization can be performed, but this method is not suitable for solving automatic history matching problem in large oil reservoir simulation

Engineering Contradiction:

Improveoptimization reliabilityVSAvoidcomputational complexity

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The system dynamically changes optimization parameters including the choice of algorithm (gradient, gradient-free, or hybrid), population size, iteration limits, and parameter adjustment steps. These parameter changes allow the system to adapt to different reservoir scales and computational resource availability, making large-scale reservoir simulation feasible.

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The patent implements dynamic optimization where the algorithm can switch between different optimization strategies during the matching process. The system dynamically adjusts the optimization approach based on problem characteristics, combining multiple methods adaptively to handle large-scale reservoir simulations effectively.

Inventive Principle:

Principle #15Dynamics

3Productivity

If the finite storage method is used to process the automatic history matching problem, then the method has certain superiority, but there is a great limitation as this method fails to being universal to oil reservoir numerical simulators

Engineering Contradiction:

Improvematching efficiencyVSAvoiduniversality

Core Design Contradiction:

ProductivityVSAdaptability or versatility

Solution Approach 1:

The patent creates a universal automatic history matching system that can interface with multiple different oil reservoir numerical simulators. The system uses standardized data exchange formats and configurable simulator interfaces, allowing the same optimization framework to work with various simulator types (finite difference, finite element, etc.) without method-specific limitations.

Inventive Principle:

Principle #6Universality (Multi-functionality)

4Adaptability or versatility

If random algorithms are applied to solving history matching problems of the oil reservoir of relatively large scales, then the algorithms can be executed, but they cannot gain satisfactory calculating efficiency and precision

Engineering Contradiction:

Improveapplicability to large scalesVSAvoidcalculating precision

Core Design Contradiction:

Adaptability or versatilityVSMeasurement precision

Solution Approach 1:

The patent merges multiple optimization approaches into a hybrid algorithm that combines gradient-based methods (for precision) with gradient-free or random search methods (for exploration). This combination allows the system to maintain high precision while handling large-scale reservoir problems, overcoming the limitations of purely random algorithms.

Inventive Principle:

Principle #5Merging (Combining)

Data Source

PatentUS12560069B2Automatic history matching system and method for an oil reservoir based on transfer learning

Publication Date: 2026.02.24 CHINA UNIV OF PETROLEUM (EAST CHINA)

AI summary

The present invention relates to an automatic history matching system for an oil reservoir based on transfer learning, comprising a data reading module, a population reinitializing module, an optimization module, a simulated calculation module, a comparative judgment module and an output module, wherein the data reading module reads an optimized result of an existing oil reservoir, outputs the optimized result to the population reinitializing module, obtains an initial population of a new oil reservoir by calculation and outputs the initial population to the optimization module.