We developed the numerical modeling of Marine Controlled Source Electromagnetic
(MCSEM) synthetic data used in hydrocarbon exploration for three-dimensional models
using parallel computation. The models are formed of two strati ed layers: the sea and
the host with a thin three-dimensional embedded reservoir overlapped by the air half-space.
In this work we present a three-dimensional nite elements technique of MCSEM modeling
using the primary and secondary decomposition of the magnetic and electric coupled
potentials. The electromagnetic elds are calculated by numerical di erentiation of the
scattered potentials. We explore the parallelism of the MCSEM 3D data in a multitransmitter
survey, where as for each transmitter position we have the same forward model but with
di erent data. For this, we use Message Passing Interface library (MPI) and the client server
approach, where the server processor sends the input data to client processors to perform
the calculation. The input data are formed by the parameters of the nite element mesh,
together with informations about the transmitters and the geoeletric model of hydrocarbon
reservoir with prismatic geometry. We observe that when the horizontal width and the length
of the reservoir have the same order of magnitude, the in-line responses are very similar and
the consequently the three-dimensional e ect is not detectable. On the other hand, when
the di erence in the sizes of the horizontal width and the length of the reservoir is very
large, the e ect 3D is easily detected in in-line along the biggest dimension of the reservoir.
For measures done along the lesser dimension this e ect is not detectable, therefore, the
model 3D approaches to a bidimensional model. The parallelism of multiple data has fast
implementation and processing, and its time of execution is of the same order of the serial
problem, with the addition of the latency time in the data transmission among the cluster
nodes, which justifying this methodology in modeling and interpretation MCSEM data. Only
simple 3D models were computed because of the reduced memory (2 Gbytes in each node)
of the cluster of UFPA Departament of Geophysics.
|
