Reservoir Well Target Identification Using Opportunity Index Embeddings
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current well target identification in reservoir simulation is labor-intensive, time-consuming, and lacks effective knowledge transfer when experts leave an organization, making it difficult to automate and scale the process.
Innovation Solution
An automated system and method that utilizes expert knowledge to identify well targets through a data-based model, enabling continuous improvement and real-time inference, using decision trees and classification models to predict well targets in reservoir simulation models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If expert knowledge is used to identify well targets, then identification accuracy is improved, but time consumption and labor intensity increase
Solution Approach 1:
The patent creates a digital twin or virtual model of the reservoir that replicates the physical reservoir's properties and behavior. This virtual model allows automated identification of well targets by simulating reservoir responses to different well placements, eliminating the need for manual expert analysis while maintaining identification accuracy through realistic simulation physics
Solution Approach 2:
The patent replaces the mechanical/manual process of expert analysis with an automated computational system. Machine learning algorithms and reservoir simulation software automatically process geological data, evaluate potential well locations, and identify optimal targets, substituting human cognitive work with computational algorithms that operate faster and without fatigue
2Reliability
If manual expert identification is used, then reasoning quality is improved, but knowledge transferability deteriorates
Solution Approach 1:
The patent captures expert knowledge by training machine learning models on examples of expert-identified well targets and their reasoning processes. The model learns to replicate expert decision-making patterns, encoding tacit knowledge into an explicit algorithm that can be transferred and applied consistently across different reservoirs and teams without relying on individual experts
Solution Approach 2:
The patent introduces a machine learning model as an intermediary that formalizes and makes explicit the implicit reasoning processes of experts. This intermediary translates subjective expert judgment into objective, documentable criteria that can be shared, reviewed, and applied consistently across the organization, improving knowledge transferability while maintaining reasoning quality
3Productivity
If automated systems are implemented, then productivity is improved, but system complexity increases
Solution Approach 1:
The patent breaks down the complex task of well target identification into distinct modular components: data preprocessing modules, reservoir simulation modules, machine learning inference modules, and result interpretation modules. Each module handles a specific aspect of the workflow independently, making the overall system more manageable, easier to maintain, and simpler to implement while maintaining high productivity through automated processing
Data Source
Figure 1A~1C
Figure 2
Figure 3
AI summary
A system and method are provided for identifying a wellsite target for drilling, including receiving a plurality of data regarding a wellsite, generating a distribution of reservoir properties using the plurality of data for an area of a reservoir defined within the wellsite, determining at least one opportunity index for an area in the reservoir based on at least one of the corresponding reservoir properties, classifying a section of the reservoir based on at least one computed embedding space, wherein the at least one computed embedding space of the section is based on the at least one opportunity index.