Use of neural networks for dynamic interpretation of seismic data
- Authors: Kaliyev D.T.1
-
Affiliations:
- KMG Engineering LLP
- Issue: Vol 4, No 2 (2022)
- Pages: 27-34
- Section: Articles
- URL: https://journals.rcsi.science/2707-4226/article/view/125888
- DOI: https://doi.org/10.54859/kjogi108576
- ID: 125888
Cite item
Full Text
Abstract
Neural networks and machine learning have long been used by almost everyone in their daily lives, perhaps not always consciously. When an algorithm of social networks identifies the faces of people in a photo or a voice assistant helps us search for some information, machine learning techniques underpin all of these activities. In recent years neural networks are finding more and more applications in the fields of oil and gas exploration and production. This article aims to illustrate an example of the application of neural networks in the analysis of seismic data for an active oilfield by predicting 3D cube of petrophysical properties to further detail the geological model and search for additional hydrocarbon accumulations. One of the key conditions for successful prediction of petrophysical properties using neural networks is a wide sample of well data for effective training of a non-linear operator. In our case, since it is a producing field, there were more than 100 wells available, which fully meets the requirements of the algorithm. Another important condition for application of this technique is having high-quality well ties for the used wells, this step of the workflow will also be described within the article. A distinct feature of neural network analysis, in contrast to classical inversion, is that it does not use a seismic wavelet. The neural network automatically determines such an operator that best describes the correlation between several seismic traces in the wellbore area and the log curve. This feature reduces the analysis time and produces express results if the above mentioned conditions are met, which makes the neural network technique an effective tool for dynamic analysis of seismic data.
Keywords
Full Text
About the authors
D. T. Kaliyev
KMG Engineering LLP
Email: d.kaliyev@niikmg.kz
руководитель группы сейсмических исследований Nur-Sultan
References
- Veeken P.C.H., Priezzhev I.I. Genetic Seismic Inversion Using a Non-linear, Multi-trace Reservoir Modeling Approach. – 71st EAGE Conference and Exhibition incorporating, SPE EUROPEC, 2009. doi: 10.3997/2214-4609.201400020.
- Priezzhev I.I., Veeken P.C.H. Seismic waveform classification based on Kohonen 3D neural networks with RGB visualization. – First Break, 2019, v. 37, iss. 2, pp. 37–43. DOI: https://doi.org/10.3997/1365-2397.2019012.
- Учебные материалы ПО Petrel от 18.05.2021. // Uchebnye materialy PO Petrel ot 18.05.2021. [Petrel Software Tutorial Materials dated 05/18/2021]
Supplementary files
