3D Permeability Model for Reservoirs Using ML Flow Unit Clustering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current reservoir characterization methods face challenges in accurately updating static and dynamic properties over the life of a field, often relying on laborious manual processes and costly invasive experiments, which can lead to inefficient data processing and biased models.
Innovation Solution
A computer-implemented method using machine learning algorithms, such as K-means clustering, to automatically detect and remove anomalous data, identify flow units, and generate a 3D permeability model that aligns with core and log data, reducing the number of geo-model grids and incorporating saturation-height functions for improved reservoir simulation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual processes and invasive experiments are used for reservoir characterization, then measurement precision can be maintained, but productivity is reduced and loss of time increases
Solution Approach 1:
The patent replaces manual mechanical processes with automated machine learning algorithms. Specifically, ML models automatically process well log data, core data, and production data to generate 3D reservoir models, eliminating the need for laborious manual interpretation and reducing processing time while maintaining characterization accuracy through advanced pattern recognition capabilities
Solution Approach 2:
The system enables self-service through automated data processing pipelines where the ML models independently analyze multiple data sources, identify patterns, and generate reservoir models without continuous human intervention. The automated workflows include data preprocessing, feature extraction, model training, and validation stages that operate autonomously to improve productivity
2Measurement precision
If manual processes and invasive experiments are used for reservoir characterization, then measurement precision can be maintained, but loss of time increases
Solution Approach 1:
The patent applies preliminary action by pre-training machine learning models on extensive datasets of well logs, core data, and production data before actual reservoir modeling. This pre-training phase enables the models to quickly process new data and generate accurate 3D models without requiring time-consuming manual analysis during the actual reservoir characterization process
Solution Approach 2:
The substitution of manual characterization methods with ML-based automated modeling significantly reduces simulation run-time. The ML models can rapidly process large volumes of data and generate 3D permeability models and saturation-height functions in minutes rather than the days or weeks required by traditional manual methods
3Manufacturing precision
If traditional methods are used for generating 3D geo-models, then manufacturing precision can be maintained, but device complexity increases
Solution Approach 1:
The patent implements universality by using a single integrated machine learning framework that handles multiple functions: processing well log data, core data, and production data; generating 3D permeability models; creating saturation-height functions; and producing flow unit classifications. This multi-functional ML system replaces multiple separate traditional tools and methods, maintaining model accuracy while simplifying the overall system architecture
Solution Approach 2:
The system utilizes parameter changes by transforming raw well log data, core data, and production data into standardized features that the ML models can process. The ML algorithms automatically adjust parameters such as permeability, porosity, and saturation based on learned patterns from training data, enabling accurate 3D model generation without complex manual parameter tuning
4Ease of operation
If traditional reservoir characterization methods are used, then ease of operation can be maintained, but productivity is reduced
Solution Approach 1:
The ML-based system provides self-service capabilities where the models automatically perform data preprocessing, feature extraction, and model generation without requiring extensive human expertise or manual intervention. The automated workflows handle the complexity internally while presenting simple inputs and outputs to users, maintaining ease of operation while dramatically improving data processing efficiency
Data Source
AI summary
Some implementations provide a method including: accessing measurement data that characterize one or more features of a reservoir, wherein the measurement data are from more than well locations of the reservoir and from a range of depths inside the reservoir; detecting portions of the measurement data that characterize the one or more features with a statistical metric that is below a pre-determined threshold; based on removing the portions of the measurement data, identifying a plurality of layers along the range of depths of the reservoir; within each layer of the plurality of layers, grouping the measurements data among a plurality of clusters, each corresponding to a flow unit (FU) and determined by a machine learning algorithm; generating a three-dimensional (3D) permeability model of the reservoir based on the FU of each layer and a saturation height function; and simulating a performance of the reservoir based on the 3D permeability model.