PERBANDINGAN ANTARA MULTI ATRIBUT SEISMIK REGRESI LINIER DAN MULTI ATRIBUT SEISMIK PROBABILISTIC NEURAL NETWORK UNTUK ESTIMASI POROSITAS RESERVOIR BATU PASIR PADA LAPANGAN MINYAK TEAPOT DOME; COMPARISON OF LINEAR REGRESSION SEISMIC MULTI ATTRIBUTE AND PROBABILISTIC NEURAL NETWORK SEISMIC MULTI ATTRIBUTE FOR SANDSTONE RESERVOIR POROSITY ESTIMATION ON THE TEAPOT DOME OIL FIELD

AZIZ, ZULFANI, Ari Setiawan

2016 | Disertasi | FMIPA

In the petroleum exploration, the informations about rocks in subsurface are very necessary to determine the reservoir zone target. One of them is the rock porosity information. To obtain the the rock porosity distribution informations, we used seismic multiattribute method that can estimate the porosity from seismic attributes. Seismic multiattribute method has two types – linear regression and probabilistic neural network (PNN). This research conducted in order to understand which method can give better result to estimate the sandstone reservoir porosity in the Teapot Dome oil field. This research used combinations from three attributes – acoustic impedance, integrate, and amplitude weighted frequency – to estimate the sandstone reservoir porosity which based on the cross validation. Linear regression seismic multiattribute assumes that the relation of three attributes and porosity is linear, while probabilistic neural network seismic multiattribute assumes non-linear. This research revealed that seismic multiattribute linear regression gives porosity estimation with 0.701 correlation value and 0.649 validatin value, whereas seismic multiattribute probabilistic neural network gives better porosity estimation with 0.920 correlation value and 0.683 validation value. The other result also showed that the porosity estimation curve from probabilistic neural network is more compatible with the original porosity curve than the porosity estimation curve from linear regression.

Kata Kunci : N