Speaker
Description
Reasonable prediction and evaluation of reservoir parameters are fundamental to reservoir geological modeling and refined hydrocarbon reservoir assessment. To address the limitations of traditional physical testing methods—such as prolonged parameter acquisition time and uncertainty in data processing, 17 artificial intelligence (AI)-based parameter prediction models were firstly compared. Through this comparative analysis, techniques including deep neural networks, support vector machines, and clustering analysis were systematically selected and applied. Furthermore, data optimization, dimensionality reduction, and decision algorithm refinement were respectively focused on in this paper, ultimately leading to the establishment of an AI-based comprehensive evaluation and optimization model along with a Python algorithmic workflow for low-permeable and tight oil reservoirs. The integrated approach was developed to enable intelligent and high-precision predictions of key petrophysical parameters, including microscopic throat radius and movable fluid content. Validation results based on optimized datasets demonstrated the effectiveness of the proposed methodology, achieving prediction accuracies of 86.25% for average throat radius, 89.9% for movable fluid percentage, 89.4% for centrifuged movable fluid percentage, and 84.7% for threshold pressure gradient in the studied low-permeable and tight oil reservoir block. The findings from the study significantly improved the accuracy of predicting fundamental petrophysical parameters. Meanwhile, it provided both a methodological framework and technical support for deepening the understanding of microstructural characteristics, development potential, and operational challenges in low-permeable and tight oil reservoirs.
| Country | China |
|---|---|
| Acceptance of the Terms & Conditions | Click here to agree |
