Naslov
Improved neural network model of assessment for interpretation of Miocene lithofacies in the Vukovar Formation, Northern Croatia

Autori
Varenina, Andrija ; Malvić, Tomislav ; Mate, Režić

Izvornik
RMZ - Materials and geoenvironment (1408-7073) 65 (2018), 3; 145-156

Vrsta, podvrsta i kategorija rada
Radovi u časopisima, članak, znanstveni

Ključne riječi
neural networks, Ladislavci Field, Drava Depression, Miocene, Croatia

Sažetak
The Ladislavci Field (oil and gas reservoirs) is located 40 km from city of Osijek, Croatia. The oil reservoir is in structural- stratigraphic trap and Miocene rocks of the Vukovar formation (informal named as El, F1a and F1b). The shalower gas reservoir is of Pliocene age, i.e. part of the Osijek Sandstones (informal named as B). The oil reservoirs consist of limestones, breccias and conglomerates, and gas is accumulated in sandstones. Using neural networks it was possbile to interpreted applicability of neural algorithm in well log analyses as well as using neural model for reservoir prediction without or with small number of log data. Neural networks are trained on the data from two wells (A and B), collected from the interval starting with border of Sarmatian/Lower Pannonian (EL marker Rs7) to the well’s bottom. The inputs were data from spontaneous potential (SP), and resistivity (R16 and R64) logs. They are used for neural training and validation as well as for final prediction of lithological composition in analysed field. The multilayer perceptron network (MLP) had been selected as the most appropriate.

Izvorni jezik
Engleski

Znanstvena područja
Geologija

POVEZANOST RADA