نویسندگان
1 کارشناسی ارشد، شرکت نفت و گاز پارس، تهران
2 استادیار، دانشکده فنی و مهندسی، دانشگاه بینالمللی امام خمینی(ره)، قزوین
3 استاد، مؤسسه ژئوفیزیک دانشگاه تهران، تهران
چکیده
کلیدواژهها
عنوان مقاله [English]
نویسندگان [English]
Seismic facies analysis is a powerful technique used in extracting stratigraphic information and reservoir properties from seismic data. This technique can be implemented as a useful tool in different stages of exploration, production and development of hydrocarbon fields. In this paper, travel time samples (seismic traces) related to seismic reflections is grouped in similar classes. Using multiple seismic attributes within the target interval help us to construct a conceptual multi-attribute space in which each attribute is considered to be related to a property, or response of rock (and/or fluid). In this study, two unsupervised and supervised techniques are implemented for seismic facies analysis. In unsupervised approach, which is based entirely on the internal structure of data, and does not use any type of well and geological data as an auxiliary input to the predicting network, a self-organizing neural network (NN) is used. The results of both classifications in the study area are highly convergent and also show good consistency with petrophysical well data. In addition to estimating the petrophysical properties in interwell spaces, the use of index wells for supervision on classifications verifies the unsupervised classification in reservoir interval. This brings about a high confidence for the application of unsupervised classification. Consequently, generation of unsupervised facies maps in different time horizons facilitates the interpretation of property changes related to each seismic facies. Reservoir properties changes related to each facies in a facies map with three classes highly correspond with the petrophysical properties concerned.
کلیدواژهها [English]