Reservoir Property Modeling with Decision Forest Envelopes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current reservoir modeling methods using geostatistics fail to accurately characterize reservoir boundaries and detect additional reservoirs, leading to uncertainties that impact the accuracy of production simulation operations.
Innovation Solution
Combining geostatistics with a decision forest obtained using machine learning (ML) to create a three-dimensional reservoir property model, which includes training a decision tree ensemble and executing a conditional simulation using intervention variables and powers to enhance accuracy in estimating subsurface properties like porosity, permeability, and initial water saturation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If geostatistics is used for reservoir modeling, then spatial prediction capability is provided, but accuracy in characterizing reservoir boundaries and detecting additional reservoirs deteriorates
Solution Approach 1:
The patent combines geostatistics with machine learning decision forest methodology to create an integrated modeling approach. The decision forest is trained on sample data including well log data, core data, and seismic data, then integrated with geostatistical simulation to generate ensemble realizations that improve both boundary characterization accuracy and reservoir detection reliability simultaneously
2Loss of information
If traditional reservoir property models are created using geostatistics, then spatial distribution of properties is estimated, but uncertainty in production simulation increases
Solution Approach 1:
The patent implements a feedback mechanism where the decision forest model is trained on historical well data and production data, then used to generate predictions that are compared against actual measurements. The model is iteratively refined using this feedback to reduce uncertainty and improve precision of property estimates for production simulation
Solution Approach 2:
The patent replaces traditional geostatistical mechanics with a machine learning decision forest approach. Instead of relying solely on statistical assumptions and random function theory, the system uses tree-based ensemble methods that can capture non-linear relationships and interactions between variables, reducing uncertainty while maintaining computational efficiency
3Reliability
If multiple reservoir property models are created and evaluated, then best models are selected for simulation, but computational time and complexity increase
Solution Approach 1:
The patent performs preliminary action by training the decision forest model on comprehensive sample data before actual reservoir modeling. The pre-trained model captures key relationships and patterns, enabling faster and more accurate property prediction during the actual modeling phase, reducing both time and computational complexity while maintaining high model quality
Data Source
AI summary
Systems and methods herein include a method for reservoir property modeling, comprising training an ML decision tree ensemble on sample data, to generate a trained decision forest; computing, by the trained decision forest, an envelope for target variables at one or more target locations based on at least one variable; and executing a conditional simulation by using the at least one variable to compute a target variable of the target variables by sampling from the envelope based on a sampling variable.