Borehole Holdup Prediction Using Machine Learning and PNL Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional methods for determining borehole fluid holdup using pulsed neutron logging (PNL) data and rule-based calculations result in low-accuracy predictions, necessitating a more accurate approach for fluid profiling and saturation corrections.
Innovation Solution
A machine learning approach is employed, utilizing simulated PNL data and lab data to train a learning machine, which generates features from energy spectra detected by near and far detectors to predict borehole holdup, with calibration curves used to convert simulated data into lab-equivalent synthetic data for improved accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If rule-based calculations are used to determine borehole holdup from PNL data, then the method is simple and easy to implement, but the prediction accuracy is low
Solution Approach 1:
The patent replaces traditional rule-based calculation methods with a machine learning model that uses neural networks to predict borehole holdup. The ML model processes PNL data, formation density, and fluid properties to generate accurate holdup predictions, substituting simple arithmetic rules with a sophisticated computational system that learns optimal relationships from training data.
Solution Approach 2:
The patent transforms the approach by changing from fixed rule-based parameters to dynamic machine learning parameters. The system uses multiple input parameters including PNL data, formation density, fluid density, and fluid composition, which are processed through the ML model to produce holdup predictions. This parameter transformation enables the system to adapt to different well conditions and achieve higher accuracy.
2Measurement precision
If machine learning models are trained with limited data, then training is faster and requires fewer resources, but model accuracy and generalization capability are reduced
Solution Approach 1:
The patent applies preliminary action by pre-training the machine learning model using synthetic data generated from Monte Carlo simulations before deploying it with real field data. This pre-training phase allows the model to learn fundamental relationships and patterns without requiring extensive real-world data, improving its generalization capability when applied to actual borehole measurements.
Solution Approach 2:
The patent uses synthetic simulation data as an intermediary between theoretical models and real field data. The Monte Carlo-generated synthetic data serves as a bridge, allowing the ML model to be trained on realistic but controlled data before being validated and deployed with actual PNL measurements, thereby reducing the need for large volumes of real training data.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
The machine learning model achieves higher accuracy in borehole holdup predictions, providing insights into fluid flow profiles and enabling more precise saturation corrections by leveraging expanded variable spaces and feature selection techniques.
Implementation Method 1
When neutrons emitted from a neutron generator on the PNL tool interact with atomic nuclei of elements present in the borehole region and subsurface formation, secondary radiation is produced
Implementation Method 2
The secondary radiation may be detected by the far and near detectors
Data Source
AI summary
In some implementations, a method for controlling a learning machine to predict fluid holdup in a borehole comprises generating an expanded dataset of simulated pulsed neutron logging (PNL) data based, at least in part, on an original dataset of empirical PNL data, converting, using one or more calibration coefficients, the simulated PNL data into lab-equivalent synthetic data, and training an ensemble of machine learning models based on the lab-equivalent synthetic data.