Machine Learning Model for Reservoir Fluid Property Estimation
Solution Overview
Problem
Conventional mud-gas logging techniques require advanced mud-gas data for accurate reservoir fluid property predictions, which is not always available, limiting the ability to predict fluid properties like gas-oil ratio and density without interrupting the drilling process or incurring additional costs.
Innovation Solution
A method using a machine learning model trained with standard mud-gas data and petrophysical data to predict reservoir fluid properties, allowing for continuous logging and reduced costs by eliminating the need for advanced mud-gas data collection, and enabling predictions of properties influenced by oil-related components without downhole fluid sampling.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If advanced mud-gas data is used for accurate reservoir fluid property predictions, then measurement precision is improved, but device complexity and cost increase
Solution Approach 1:
The patent creates a virtual copy of advanced mud-gas data by training a machine learning model on standard mud-gas data. The model learns to replicate the predictive capabilities of advanced systems using only standard equipment data, eliminating the need for complex advanced sensors while maintaining prediction accuracy
Solution Approach 2:
The patent replaces the physical advanced mud-gas logging system with a computational machine learning model. Instead of using complex hardware to directly measure fluid properties, the system uses software algorithms to infer properties from standard mud-gas data, substituting mechanical measurement with computational analysis
2Measurement precision
If downhole fluid sampling is performed for fluid property analysis, then measurement precision is improved, but productivity decreases due to drilling interruption
Solution Approach 1:
The patent enables continuous fluid property prediction by using machine learning models that process standard mud-gas data in real-time during drilling operations. This eliminates the need to stop drilling for fluid sampling, maintaining continuous productive action while providing ongoing fluid property information
Solution Approach 2:
The patent introduces machine learning models as an intermediary between standard mud-gas data and fluid property predictions. This intermediary enables indirect measurement of fluid properties through computational inference, replacing the direct but disruptive method of physical fluid sampling
3Ease of operation
If standard mud-gas data is used without recycling correction, then ease of operation is improved, but measurement precision deteriorates
Solution Approach 1:
The patent performs preliminary correction of standard mud-gas data by applying recycling corrections during the data collection phase. This advance correction ensures that the data fed into the machine learning model is already accurate, eliminating the need for post-processing corrections and maintaining both simplicity and precision
Data Source
AI summary
A method is disclosed for generating a machine learning model to predict a reservoir fluid property, such as gas-oil ratio or density, based on standard mud-gas and petrophysical data. It has been found that this model predicts these reservoir fluid properties with an accuracy that is close to that which can be achieved using advanced mud-gas data. This is advantageous, as than standard mud-gas data and petrophysical data is much more readily available than advanced mud-gas data.