Prediction of High-Pressure Physical Properties of Crude Oil Using Explainable Machine Learning Models
High-pressure physical property parameters of formation crude oil are crucial for oilfield exploration, development, and production. Various prediction models have been developed using PVT experimental methods, empirical formulas, regression analysis, and machine learning techniques. However, these...
Saved in:
| Main Authors: | , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2025-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10815941/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | High-pressure physical property parameters of formation crude oil are crucial for oilfield exploration, development, and production. Various prediction models have been developed using PVT experimental methods, empirical formulas, regression analysis, and machine learning techniques. However, these models often lack sufficient transparency and clarity in their computational processes, limiting their effectiveness in oil and gas exploration and development. This paper employs the Extreme Gradient Boosting (XGBoost) machine learning algorithm, utilizing experimental data such as formation pressure, formation temperature, surface crude oil density, natural gas relative density, gas-oil ratio, saturation pressure, volume factor, formation crude oil density, and formation crude oil viscosity as the dataset. The machine learning model is trained with these parameters to develop the XGBoost prediction model. Optimal parameters, including gas-oil ratio, saturation pressure, volume factor, formation crude oil density, and formation crude oil viscosity, are selected, and the Particle Swarm Optimization (PSO) algorithm is applied to further optimize the model, resulting in the PSO-XGBoost prediction model. To enhance the understanding of the model’s operation and the influencing factors, Shapley Additive Explanations (SHAP) and Explainable Boosting Machine (EBM) techniques are employed for both local and global interpretations of the machine learning model. This study identifies key factors and their weights in predicting high-pressure physical property parameters of formation crude oil, providing valuable insights for predicting shale oil reservoir parameters and enhancing exploration and development efficiency. |
|---|---|
| ISSN: | 2169-3536 |