TITLE:
Application of Machine Learning Methods in Parameter Prediction of Deep Lacustrine Oil Shale Reservoirs—Taking the Triassic Chang 7 in the Longdong Area as an Example
AUTHORS:
Yuxin Shen, Zhichao Liu
KEYWORDS:
Machine Learning, Ordos Basin, Porosity, Permeability, Underground Reservoir Prediction
JOURNAL NAME:
Journal of Geoscience and Environment Protection,
Vol.13 No.3,
March
13,
2025
ABSTRACT: In the course of oil and gas exploration, understanding the petrophysical parameters such as reservoir porosity and permeability is crucial for evaluating oil and gas reserves and mining effectiveness. However, due to the complexity of well logging data and the interrelation between data, traditional analysis methods often have certain limitations and deficiencies. In order to overthrow the constraints of traditional methods, machine learning algorithms are used to build prediction models for porosity and permeability. In this study, Chang 7 reservoir in Longdong area was taken as the research object, and 320 core samples were selected for detailed measurement of porosity and permeability. In order to establish a generative prediction model, four different machine learning methods were used, including random forest (RF), K-nearest neighbor (KNN), extreme gradient boosting tree (XGBoost) and support vector machine (SVM). These methods were used to build an accurate prediction model for porosity and permeability combined with core and well logging data. In the experimental process, data preprocessing and feature selection were carried out, and four distinct machine learning methods were utilized to train and verify the model, and the optimal algorithm was selected according to the accuracy and stability of the model. Through single well analysis, the effectiveness of machine learning-based methods in predicting porosity and permeability was verified. Machine learning methods can build efficient and accurate prediction models by deeply mining information in data, providing researchers with a more detailed and comprehensive understanding of reservoir characteristics. This technological progress not only optimizes the decision-making process of reservoir development, but also improves the effectiveness of resource application, which is of great value to the advancement of the petroleum industry.