1. From traditional extrapolation to neural networks: Time-depth relationship innovations in the subsurface characterization of Drava Basin, Pannonian Super Basin.
- Author
-
Kamenski, Ana, Cvetković, Marko, Kapuralić, Josipa, Močilac, Iva Kolenković, and Brcković, Ana
- Subjects
- *
EXTRAPOLATION , *HYDROCARBONS , *ACOUSTIC oil well logging , *ARTIFICIAL neural networks - Abstract
The estimation of time-to-depth relationships can prove challenging in regions with rare acoustic logs. This study focuses on the eastern part of the Drava Basin in north Croatia, chosen as a mature hydrocarbon exploration area with abundant geophysical and well data. As only a small portion of wells have well log measurements or seismic profiling performed, a time-to-depth extrapolation is often performed, which potentially results in the erroneous placement of well log markers in the time domain and affects the interpretation of seismic sections or volumes. This study proposes a novel methodology for predicting two-way travel time values in wells without vertical seismic profiling or acoustic logging. This research evaluates the parameters for the characterization of the velocity distribution in the subsurface and the efficiency of artificial neural networks versus conventional methods for this task. The constructed artificial neural network model has a correlation coefficient above 0.99 for the training, testing, and validation datasets, with a mean absolute error of approximately 25 milliseconds for each network. Artificial neural networks proved to have a lesser error in predicting the two-way time and are not sensitive to outlier values. [ABSTRACT FROM AUTHOR]
- Published
- 2024
- Full Text
- View/download PDF