Holdup Measurement Using Quantized Classification and Local Regression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional holdup measurement techniques, particularly pulsed neutron log-based methods, face challenges in accuracy and applicability due to fluid/gas segregation in horizontal wells and limited applicability to various borehole conditions, leading to unreliable measurements of oil saturation and flow determination.
Innovation Solution
The application of machine learning techniques during gamma spectrum analysis to classify and quantify holdup measurements, utilizing gamma spectral data to differentiate between oil, water, and gas volumes in a wellbore, thereby improving measurement accuracy and applicability across diverse borehole environments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If traditional pulsed neutron log-based methods are used for holdup measurement, then the measurement process is simplified, but the accuracy and reliability of holdup measurements deteriorate due to fluid/gas segregation in horizontal wells and limited applicability to various borehole conditions
Solution Approach 1:
The patent replaces traditional mechanical/neutron-based measurement systems with a gamma-ray spectroscopy system combined with machine learning algorithms. The gamma detector system measures gamma ray spectra from naturally occurring radioactive materials in oil, water, and gas phases, and machine learning models classify and quantify holdup based on spectral characteristics, eliminating the need for complex neutron sources and mechanical separation devices.
Solution Approach 2:
The patent transforms the measurement approach by changing from direct physical measurement to spectral analysis. Instead of measuring holdup directly through neutron interaction, the system measures gamma ray energy spectra and uses machine learning to infer holdup parameters. This parameter transformation enables accurate measurement in horizontal wells where traditional methods fail due to fluid segregation.
2Device complexity
If traditional pulsed neutron log-based methods are used for holdup measurement, then the device complexity is reduced, but the applicability to diverse borehole conditions deteriorates
Solution Approach 1:
The patent creates a universal measurement system that can handle diverse borehole conditions through machine learning models trained on multiple datasets. The gamma-ray spectral analysis system with quantized classification and categorized local regression can adapt to different well orientations (vertical, inclined, horizontal), various fluid compositions, and different borehole geometries, providing a single platform that replaces multiple specialized measurement tools.
3Measurement precision
If machine learning techniques with quantized classification and categorized local regression are applied to gamma spectrum analysis, then the accuracy and precision of holdup measurements are improved, but the device complexity and computational requirements increase
Solution Approach 1:
The patent segments the continuous holdup measurement problem into quantized classification levels (e.g., low, medium, high holdup categories). The machine learning model first classifies the gamma spectrum into discrete quantized levels, then applies categorized local regression within each level to predict precise holdup values. This segmentation simplifies the computational complexity while maintaining high measurement precision across different holdup conditions.
4Measurement precision
If rigorous computer simulations are used for holdup measurement calibration, then the measurement accuracy is improved, but the time and computational resources required increase
Solution Approach 1:
The patent performs comprehensive calibration and model training in advance during an offline phase. Machine learning models are trained on extensive simulated and field data to learn the relationship between gamma spectra and holdup parameters. Once trained, the models can rapidly process field measurements without requiring real-time simulations, significantly reducing measurement time while maintaining high accuracy through the preliminary computational work.
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
This approach enhances the accuracy and generality of holdup measurements by leveraging machine learning models to analyze gamma spectral data, providing precise volumetric ratios of oil, water, and gas, even in complex horizontal well conditions, and reduces dependency on rigorous computer simulations.
Implementation Method 1
gamma detectors absorb the incident gamma energy and generate electric pulses from Compton scattering (i.e., scattering of a photon after an interaction with a charged particle), photoelectric, and pair-production mechanisms
Implementation Method 2
gamma detectors absorb the incident gamma energy and generate electric pulses from Compton scattering, photoelectric, and pair-production mechanisms
Implementation Method 3
gamma detectors absorb the incident gamma energy and generate electric pulses from Compton scattering, photoelectric, and pair-production mechanisms
Data Source
AI summary
Aspects of the subject technology relate to determining a holdup measurement based on a gamma spectrum through machine learning. A spectral image based on a gamma spectrum generated downhole in a wellbore can be accessed. A component of a holdup measurement for the wellbore can be classified into a specific quantized level through application of a machine learning classification model to the spectral image. A continuous value for the component of the holdup measurement for the wellbore can be quantified by applying a machine learning quantization model associated with the quantized level.