Reservoir Property Upscaling with Machine Learning for Borehole Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods struggle to upscale hydrocarbon reservoir rock or fluid properties from well sample scale to borehole scale accurately, leading to underestimation or overestimation of reservoir volumes due to sparse and scanty measurements, particularly for cementation factor, which affects water saturation and reservoir volume estimation.
Innovation Solution
A machine learning model, trained using archival wireline logs and corresponding cementation factor measurements, is applied to predict cementation factor values at the borehole scale, establishing a nonlinear relationship to enhance accuracy and completeness of reservoir property estimation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If core, plug, or thin section samples are analyzed in a laboratory at nanometer or micrometer scales, then ground-truth data is obtained, but the data cannot be directly integrated with borehole scale data for reservoir characterization
Solution Approach 1:
The patent applies dimensionality change by upsaling rock or fluid properties from well sample scale (nanometer or micrometer) to borehole scale through a trained machine learning model. This transforms data across different spatial dimensions, enabling integration of high-precision laboratory measurements with borehole-scale reservoir evaluation workflows.
2Ease of operation
If sparse measurements are used for reservoir characterization, then data collection is simplified, but reservoir volume estimation becomes inaccurate due to underestimation or overestimation
Solution Approach 1:
The machine learning model performs self-service by automatically learning the complex nonlinear relationships between wireline log measurements and cementation factor values. The model trains on available sparse measurements and wireline logs, then independently predicts cementation factor at borehole scale without requiring additional manual data collection or complex processing of the sparse measurements.
Solution Approach 2:
The patent changes parameters by using a trained machine learning model to predict cementation factor values from wireline log measurements. This transforms the approach from directly using sparse measurements to using learned parameter relationships, enabling accurate reservoir volume estimation while maintaining ease of operation with existing data.
3Device complexity
If traditional upscaling methods are used from well sample scale to borehole scale, then the process is simple, but accuracy is compromised due to inability to capture geological patterns
Solution Approach 1:
The patent replaces traditional mechanical or deterministic upscaling methods with a machine learning-based system. The trained model learns complex nonlinear relationships and geological patterns from training data, substituting simple interpolation or averaging methods with an intelligent system that captures subsurface heterogeneity while maintaining reasonable computational complexity.
Data Source
AI summary
Example methods and systems for upscaling rock or fluid properties of a hydrocarbon reservoir from well sample scale to borehole scale are disclosed. One example method includes obtaining one or more wireline logs of a well interval in a hydrocarbon reservoir. A trained machine learning (ML) model is applied to the one or more wireline logs to upscale one or more rock or fluid properties of the hydrocarbon reservoir to a borehole scale, where the trained ML model includes a set of weight factors, and applying the trained ML model to the one or more wireline logs includes applying the set of weight factors to the one or more wireline logs to determine the one or more rock or fluid properties at the borehole scale. The determined one or more rock or fluid properties is provided to determine a volume of the hydrocarbon reservoir for exploration of the hydrocarbon reservoir.