Pore fluid properties data-driven based on seismic parameters in porous media through supervised machine learning techniques
Pore fluid properties data-driven based on seismic parameters in porous media through supervised machine learning techniques
Esmael Makarian1 Ahmad Taheri2 Adib Amini3 Karamollah Bagherifard4
1) Department of Mining Engineering, Sahand University of Technology, Tabriz, Iran,
2) Department of Computer Engineering, School of Engineering, Yasouj University, Yasouj, Iran,
3) Department of Civil Engineering, Sahand University of Technology, Tabriz, Iran,
4) Department of Computer Engineering, Yasooj Branch, Islamic Azad University, Yasouj, Iran,
IOI Code
:
XEAK-GEEHA
Publication :
6th International Conference on Applied Researches in Science & Engineering - Germany(6carse.com)
Abstract :
This study attempts to estimate the saturation of oil and water accurately in pores in an oil reservoir as well as the electrical resistivity of the target zone based on deep learning methods. We calculated water (SW) and oil saturation (SO), and later log deep (LLD), later log Shallow (LLS), as well as Microspherically Focused Log (MSFL) using three well-known supervised machine learning techniques. In this research, we use multi-layer perceptron (MLP), Group Method of Data Handling (GMDH), and Support Vector Regression (SVR). Seismic parameters consisting of P – wave velocity (VP), S -wave data (VS), velocity ration (VP /VS), and acoustic impedance (AI) are employed as inputs data. Finally, the obtained results are compared by correlation (R), Root Mean Square Error (RMSE), and Mean Square Error (MSE). The results show that the MLP method is the best intelligent approach to estimating target parameters, especially for oil saturation with R = 0.95, RMSE: 0.6527, and MSE = 0.004.
Keywords : fluid saturation electrical resistivity machine learning seismic data porous media