Machine Learning Model for Transient Inflow Performance Relationship

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Determining the inflow performance relationship for reservoirs, especially unconventional ones like shale and tight reservoirs, is challenging due to changing conditions over time, requiring costly and time-consuming well testing or simulations.

Innovation Solution

A machine learning model is trained using type reservoir model information and production simulations to predict transient inflow performance relationships, facilitating efficient determination with acceptable accuracy and reduced costs.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If traditional well testing or production simulations are used to determine inflow performance relationship, then measurement precision is improved, but loss of time and loss of energy increase

Engineering Contradiction:

Improveinflow performance relationship determination accuracyVSAvoidtime required for well testing or simulations

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent creates a virtual copy of the reservoir system through a machine learning model trained on production simulation data. This digital replica allows repeated inflow performance relationship determinations without physical well testing, achieving measurement precision while eliminating time loss associated with actual field tests.

Inventive Principle:

Principle #26Copying

Solution Approach 2:

The machine learning model is pre-trained using production simulation data before actual inflow performance determination is needed. This preliminary training allows the system to quickly determine inflow performance relationships for new scenarios without performing time-consuming well testing each time.

Inventive Principle:

Principle #10Preliminary action

2Measurement precision

If traditional well testing or production simulations are used to determine inflow performance relationship, then measurement precision is improved, but use of energy increases

Engineering Contradiction:

Improveinflow performance relationship determination accuracyVSAvoidenergy consumed by well testing or simulations

Core Design Contradiction:

Measurement precisionVSUse of energy by moving object

Solution Approach 1:

By replacing energy-intensive physical well testing with a computational machine learning model, the patent eliminates the energy consumption associated with drilling, equipment operation, and data collection while maintaining measurement precision through trained predictive algorithms.

Inventive Principle:

Principle #26Copying

Solution Approach 2:

The patent substitutes the mechanical well testing process with an information-based machine learning system. Instead of physically measuring reservoir performance through drilling and equipment operation, the system uses trained neural networks to predict inflow performance relationships, dramatically reducing energy requirements.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

3Adaptability or versatility

If multiple production simulations are generated for different input parameter values, then adaptability is improved, but device complexity increases

Engineering Contradiction:

Improveability to handle different production scenariosVSAvoidcomplexity of production simulation system

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent creates a simplified computational copy of the complex reservoir system in the form of a machine learning model. This model captures the essential relationships between input parameters and production outcomes, providing adaptability to different scenarios while reducing the complexity of running multiple detailed production simulations.

Inventive Principle:

Principle #26Copying

Solution Approach 2:

The patent uses a machine learning model that can efficiently handle parameter variations by learning the relationships between input parameters and production outcomes during training. This allows the system to adapt to different production scenarios through parameter changes without requiring complex re-simulation procedures.

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS20230272703A1Workflow of inflow performance relationship for a reservoir using machine learning technique

Publication Date: 2023.08.31 CHEVRON USA INC

AI summary

A machine learning model is trained to facilitate determination of transient inflow performance relationship for a reservoir. A type reservoir model for the reservoir is developed and run multiple times with different input parameters to generate multiple production simulations for the reservoir. The input parameters and the results of the multiple production simulations for the reservoir are used to train a machine learning model. The trained machine learning model facilitates determination of transient inflow performance relationship for the reservoir by providing time-series prediction of average pressure, production rate, and absolute open flow of the reservoir.