Thin-layer detection using spectral inversion and a genetic algorithm

Kelyn

Description

ABSTRACT
Spectral inversion using a genetic algorithm (gA) as an optimisation approach was used for increasing the seismic
resolution of a particular dataset
/8. Al usarse
en un volumen de la formación Barco en la Cuenca del Catatumbo, Colombia, la inversión espectral recuperó
información que contribuyó a la identificación de delgados canales distributarios depositados en un ambiente
transicional con influencia de mareas. El resultado mostró que esta estrategia es más eficiente y efectiva que los
procedimientos convencionales de minimización.

Sujets

• Genetic algorithm
• Seismic inversion

Informations

• Publié par: erevistas
• Publié le: 01 janvier 2011
• Nombre de lectures: 21
• Langue: English
• Poids de l'ouvrage: 4 Mo

Extrait

EARTH SCIENCES
RESEARCH JOURNAL
Earth Sci. Res. SJ. Vol. 15, No. 2 (December, 2011): 121 -128
SEISmIC PROCESSINg
Tin-layer detection using spectral inversion and a genetic algorithm
1 2 1Kelyn Paola Castaño , g ermán Ojeda and Luis montes
1 Universidad Nacional de Colombia, Facultad de Ciencias, Departamento de g eociencias, Sede Bogotá, Colombia.
E-mail: [e-mail], [e-mail]
2 ECOPETROL S.A. Bogotá, Colombia. E-mail: [e-mail]
ABSTRACT
Keywords: genetic algorithm, spectral inversion, seismic in-
version, Barco formation, Catatumbo basin, seismic resolu-
tion.Spectral inversion using a genetic algorithm (gA) as an optimisation approach was used for increasing the seismic
resolution of a particular dataset; by contrast with the conjugate gradient method, a gA does not require a good
starting model but rather a search space.
Te method discriminated layers thinner than λ/8 when tested on synthetic and log data. When applied to
a seismic dataset concerning the Barco formation in the Catatumbo basin, Colombia, spectral inversion led to
recovering information from seismic data contributing towards the vertical identifcation of geological features
such as thin distributary channels deposited in a deltaic environment having a tidal infuence. Te results revealed
that a gA outperformed traditional minimisation schemes.
RESUmEN
Palabras claves: algoritmos genetico, inversion spectral, inver-
sion sismica, formacion barco, Cuenca Catatumbo, resolu-
cion sísmica.La inversión espectral, a partir de un algoritmo genético fue usado como estrategia de optimización, e incremento
de la resolución de un volumen sísmico. Contrario al método del gradiente conjugado, el algoritmo genético no
requiere un buen modelo inicial sino un espacio de búsqueda.
Usado en datos sintéticos y registros de pozo el método discriminó capas tan delgadas como λ/8. Al usarse
en un volumen de la formación Barco en la Cuenca del Catatumbo, Colombia, la inversión espectral recuperó Record
información que contribuyó a la identifcación de delgados canales distributarios depositados en un ambiente
transicional con infuencia de mareas. El resultado mostró que esta estrategia es más efciente y efectiva que los manuscript received: 30/04/2011
procedimientos convencionales de minimización. Accepted for publications: 10/12/2011
Introduction becoming a powerful technique for mapping thickness and geological dis-
continuity. A set of seismic attributes based on the spectral analysis of seismic
Te reliable estimation of thickness during exploration activities can refection was drawn up to map changes in thin reservoirs (marfurt and Kir-
help to delineate prospects despite a lack of well information; it may help lin, 2001). Spectral decomposition has been successfully used for enhancing
in obtaining more accurate volumetric hydrocarbon production calcula- stratigraphic features. Portniaguine and Castagna (2004) introduced a meth-
tions and in planning drilling involving lower risk. od for spectrally decomposing a seismic trace, solving the inverse problem by
Vertical seismic resolution is the key to extracting detailed stratigraphy the conjugate gradient to minimise the objective function; this technique is
which can distinguish two close seismic events associated with geological computationally time-consuming due to the method’s iterative nature. Spec-
events. Important oil reservoirs often have layers having a thickness which tral inversion uses interference patterns for identifying and characterising
is below seismic resolution. Te transitional Palaeocene Barco formation in layers (Partyka, 2005); other results have shown that inverting seismic data
the Catatumbo Basin consists of varying thickness sand wedges and, as it through spectral decomposition provides refectivity profles which are then
has several producing oil felds, there is great interest in mapping thin layers used for estimating layer thickness (Puryear and Castagna, 2008). Seismic
within it for drilling future wells. According to the Rayleigh criterion, lay- data’s vertical resolution has been increased by using the spectral inversion
ers thinner than λ/8 cannot be resolved by seismic imaging (Kallweit and method, examining the thin-bed tuning response in the frequency domain.
Wood, 1982; Widess, 1973); tuning-thickness analysis based on this model Using a genetic algorithm (gA) meant that spectral inversion gave a global
has been used as a thickness mapping method for several decades now. solution by minimising the objective function.
Amplitude and frequency variations through layers having changing Te method was tested on synthetic and log data and applied to seis-
thickness was used as a tool for extracting stratigraphic details (Partyka et mic data for interpreting thin distributary channels deposited in a deltaic
al., 1999), spectral decomposition using discrete Fourier transform (DFT) environment having a tidal infuence. 122 Kelyn Paola Castaño, g ermán Ojeda and Luis montes
Geological setting of two domes. A main fault and secondary fault to the east of the former
separate these two domes’ structures (Figure 1A). Wells P33, P34 and P35
Te Barco formation consists of interbedded fne-grained sandstone reach the top of the Barco formation on the south dome. Tis forma-
with mudstone; it was deposited in a deltaic environment having a tidal tion is divided into four depositional cycles, each representing a maximum
infuence (Nuñez and Saavedra, 2006). Tin sequences of bioturbated fooding surface and hence stratigraphic limits. Figure 1B shows x-line
fne-grained sandstone are cross bedded with fne-grained sandstone and 183 where the Barco formation and the four depositional cycles link wells
gray mudstone. Some coal horizons are present, mainly at the top of this P34 and P35.
unit. Te formation’s thickness varies from 500 feet to 700 feet, running
northeast to southeast of the basin. Teory
Two folds have been interpreted in the area (Ojeda et al., 2009);
a syncline has been located in the western part by the 3D seismic pro- An appropriate seismic attribute must be directly sensitive to geo-
gramme, followed by an anticline structure in the eastern part consisting logical features or reservoir properties, allowing a geologist to interpret a
(a)
(b)
Figure 1. (a) Top: structural contour map of the Barco formation (depth in feet); the prospect is of an anticline formed by two domes. (b) Te 183 x-line shows the two
domes and wells P34 and P35 reaching the top of the south dome and the four depositional cycles of the Barco formation.Tin-layer detection using spectral inversion and a genetic algorithm 123
geological structure and its particular environment. Coherence integrates volution between r(t) and a known wavelet w(t), the spectral decomposi-
information contained in adjacent traces to extract information which tion of a seismic trace s(t) within time window length tw can be expressed
may not be easily recognised on time scale maps; high coherence thus in- as:
dicates lateral lithological continuity whereas abrupt changes may suggest
faults and fractures (Chopra & m arfurt, 2007). Fault-to-fault coherence (7)
is used to enhance stratigraphic and structural discontinuity, discerning
faults and channel geometry. Root mean square (RmS) amplitude may A too short window afects frequency resolution and a too long win-
be applied to recognise seismic anomalies and squaring amplitude values dow deteriorates time resolution. When the wavelet spectrum is known,
within an analysis window can enhance important amplitudes above noise r(t) and T(t) are estimated by optimising the function:
level. On other hand, combination of attributes has been used to charac-
terize reservoirs (g uerrero et al, 2010) and has allowed identify channels
in transitional environment with tidal infuence.
Spectral inversion
(8)
Due to the Widess model occurring in nature as the exception not the
rule, spectral inversion is based on the fact that an impulse pair of refec-
tors can be decomposed in odd and even parts (Castagna, 2004; Chopra et
al., 2006). Seismic resolution is increased because the odd part construc- Here, f and f are high cut-of and low cut-of frequency, α and α
H L e o
tively interferes when thickness becomes thinner. are weighting functions. Te α/α ratio is adjusted according an accept-
o e
Te process difers from conventional methods because inversion is able trade-of between noise and resolution and α >> α in the case of the
o e
driven by geological knowledge and is based on local frequency spectrum Widess model.
characteristics obtained from spectral decomposition techniques (marfurt Te best solution is achieved when objective function O (t, r, r , T),
e o
and Kirlin, 2001). given by equation (6), is minimised over the frequency rank, supplying
Inversion is defned in this method (Puryear and Castagna, 2008) model parameters r , r and T.
e o
through an amplitude spectrum’s constant periodicity for a layer of given
thickness; it takes advantage of the fact that spacing between spectral peaks Genetic algorithm
and notches is precisely the inverse of layer thickness in the time domain
(Partyka et al., 1999). Unlike local op