Naslov
Neural network prediction of the reservoir properties on the Klostar oil field

Autori
Cvetković, Marko ; Velić, Josipa ; Malvić, Tomislav

Vrsta, podvrsta i kategorija rada
Sažeci sa skupova, neobjavljeni rad, znanstveni

Skup
XII. CONGRES OF HUNGARIAN GEOMATHEMATICS AND THE FIRST CONGRESS OF CROATIAN AND HUNGARIAN GEOMATHEMATICS

Mjesto i datum
Mórahalom, Mađarska, 29.05.2008. - 31.05.2008

Vrsta sudjelovanja
Pozvano predavanje

Vrsta recenzije
Nije recenziran

Ključne riječi
neural network; Kloštar field; prediction; lithology; saturation; Sava depression; Pannonian basin; Croatia

Sažetak
Oil field Kloštar is situated in the northern part of the Sava depression within the Croatian part of the Panonian basin. Main petroleum reserves are in Late Miocene sandstones which are grouped in two operative units – “ 1st and 2nd sandstone series” . Well log data from two wells (A and B) were used for the neural network analysis. Selected intervals on the two well logs correspond to the previous mentioned units. In the first analysis input data were spontaneous potential and resistivity logs (R16 and R64) while the predicted value was lithology described with categorical values as either sandstone (1) or marl (0). Training and prediction phase of the analysis were made on the same well log but on the different well log intervals, e.g. the training was done on the “ 1st sandstone series” while the prediction was done on the “ 2nd sandstone series” and vice versa. Secondly, the neural network was used to predict hydrocarbon saturation. Neural network was trained on the well A and prediction was done on the well B. Input data was extended with corresponding well log data depth (m) and the lithology. In the lithology prediction part of the study radial basis function and multi layer perceptron neural networks were used while in the hydrocarbon saturation prediction only multi layer perceptron network was used. Relatively small prediction error values and very good correspondence between predicted and real values was achieved. This points out to great possibilities in neural network application on petroleum geology problems and in exploration. At present, this study represents the best application of neural networks in the Croatian part of the Pannonian basin.

Izvorni jezik
Engleski

Znanstvena područja
Matematika, Geologija

POVEZANOST RADA