Paulo Johann

Unsupervised seismic facies analysis using wavelet transform and self-organizing maps

Unsupervised seismic facies analysis provides an effective way to estimate reservoir properties by combining different seismic attributes through pattern recognition algorithms. However, without consistent geological information, parameters such as the number of facies and even the input seismic attributes are usually chosen in an empirical way. In this context, we propose two new semiautomatic alternative methods. In the first one, we use the clustering of the Kohonen self-organizing maps (SOMs) as a new way to build seismic facies maps and to estimate the number of seismic facies. In the second method, we use wavelet transforms to identify seismic trace singularities in each geologically oriented segment, and then we build the seismic facies map using the clustering of the SOM. We tested both methods using synthetic and real seismic data from the Namorado deepwater giant oilfield in Campos Basin, offshore Brazil. The results confirm that we can estimate the appropriate number of seismic facies through the clustering of the SOM. We also showed that we can improve the seismic facies analysis by using trace singularities detected by the wavelet transform technique. This workflow presents the advantage of being less sensitive to horizon interpretation errors, thus resulting in an improved seismic facies analysis.

Campos Basin: Reservoir Characterization and Management - Historical Overview and Future Challenges

The first oil discovery in the Campos Basin dates from 1974, when the ninth well drilled found Albian carbonate reservoirs (Garoupa Field) under a water depth of 120 m. Oil production started on August 13, 1977, from the Enchova Field, which produced to a semi submersible platform moored at a water depth of 124 m. This was the beginning of a successful history that led Petrobras to become a world leader company in petroleum exploration and production in deep and ultra-deep waters. Forty-one oilfields were found between 50 and 140 km off the Brazilian coast (under water depths between 80 and 2,400 m), which produce from a variety of reservoirs, including Neocomian fractured basalts, Barremian coquinas, early Albian calcarenites, and (mostly) late Albian to early Miocene siliciclastic turbidites. These reservoirs were responsible for an average oil production of 1.2 million bpd in the year 2002 (83% of the total Brazilian production), and they are expected to be producing 1.6 million bopd by the end of 2005. The cumulative oil production from the Campos Basin comprises 3.9 billion bbl, and the current proven oil reserves are 8.5 billion bbl (89% of total Brazilian reserves). Deep and ultra-deep water giant fields started to be discovered only in 1984. There was a succession of large discoveries, including Albacora, Marlim, Albacora Leste, Marlim Sul, Barracuda, Caratinga, Roncador and, more recently, Jubarte and Cachalote. The development of these fields has continuously provided new challenges for the reservoir characterization and management in the Campos Basin. These fields are developed with fewer, horizontal and high angle wells, drilled into poorly consolidated reservoirs. The extensive use of 3D seismic as a reservoir characterization tool has optimized well location and allowed the reduction of geological risks. Integration of high-resolution stratigraphic analysis with 3D seismic inversion, geostatistic (stochastic) simulation of reservoir properties constrained by seismic, well log and core data, 3D visualization, and voxel-based automatic interpretation has guided the positioning of horizontal wells through thin (<10-15 m) reservoirs. Additionally, 3D visualization techniques have provided a new environment for teamwork, where seismic, well log, and core data are interpreted and added to detailed 3D geological models and, subsequently, to robust reservoir simulation models. The deepwater subsea wells must be designed to allow high production rates (typically >10,000-15,000 bopd), with lifetime completions to avoid costly interventions. In order to assure high productivity, pressure maintenance must be efficient; if water injection is planned, the hydraulic connectivity between injector and producer wells must be guaranteed by high-quality 3D seismic, well log correlation, and observed pressure profiles. Detailed studies have been made in order to define the distribution and number of wells, since the number of wells strongly affects the net present value of deepwater projects. Wells with expected oil recovery of less than 10-15 million bbl are not drilled in the beginning of the projects, and remain as future opportunities to increase oil production and recovery. Some of the new technologies devised for the characterization and development of the deepwater oilfields from the Campos Basin include reservoir imaging with prestack, depth-migrated seismic, 4D seismic, real-time well steering and updating of geological/reservoir models, extended reach wells, selective completion in gravel-packed wells, isolation inside horizontal, gravel-packed wells, intelligent completion, subsea oil-water separation, re-injection of produced water, scale prevention and treatment, and improved recovery techniques for heavy and/or viscous oil. Introduction Campos Basin is located in southeastern Brazil, mostly offshore of the states of Rio de Janeiro and Espírito Santo, occupying an area of 115,000 km (Fig. 1). The basin has a small (500 km) onshore portion, where the first exploratory well was drilled in 1959; this well records a 1,690 m-thick, very sand-rich Tertiary succession, Neocomian basalts, and the Precambrian metamorphic basement. Exploration in the offshore Campos Basin started in 1968, with the acquisition of 2D seismic data. The first offshore well was drilled in 1971. The first oil discovery dates from 1974, when the ninth well drilled found Albian carbonate reservoirs (Garoupa Field) at a water depth of 120 m. Oil production started on August 13, OTC 15220 Campos Basin: Reservoir Characterization and Management – Historical Overview and Future Challenges Carlos H.L. Bruhn, José Adilson T. Gomes, Cesar Del Lucchese Jr., and Paulo R.S. Johann / Petrobras E&P, Rio de Janeiro, Brazil

Wavelet transform Teager-Kaiser energy applied to a carbonate field in Brazil

Spectral decomposition has prov-en a powerful means to identify strong amplitude anomalies at specific frequencies that are otherwise buried in the broadband response. Partyka et al. (1999) showed that the seismic spectrum response from a short time window depends on the acoustic properties and thickness of the layers spanned by the window. They applied this idea to good quality marine data to delineate thin channels in Tertiary sediments in the Gulf of Mexico. They also applied spectral decomposition to moderate quality land data to delineate incised channels in Paleozoic rocks in the U.S. midcontinent. Since then, spectral decomposition has been applied to reservoir characterization, hydrocarbon detection, and stratigraphic analysis.

4D seismic in a heavy-oil, turbidite reservoir offshore Brazil

The challenge of this deepwater 4D project was to acquire seismic data over 1520 km2 in a heavily obstructed oil field. Currently, the Marlim Complex (Marlim, Marlim Sul, Marlim Leste, and Voador oil fields) produces more than 550,000 b/d from 10 fixed production platforms.

Reservoir Geophysics: Seismic Pattern Recognition Applied to Ultra-Deepwater Oilfield in Campos Basin, Offshore Brazil

Usually, seismic data is used in a qualitative approach to detect changes in the waveform and to pick acoustic continuity of a peak and/or a through as a structural mapping tool. The seismic interpretation is a qualitative process for building a geological model. Today, many works try to use the seismic information in a quantitative approach. Seismic interpretation in a quantitative approach is a key process in the integration of geoscience data at scales from basin-wide studies, reservoir focused and field-development process. Quantitative modeling could be deterministic and/or probabilistic. We use, in many steps of a seismic processing sequence, examples of quantitative deterministic modeling like seismic migration, some seismic inversion methodology, etc. Probabilistic modeling can be gathered in two groups: multivariate statistics and geostatistics approaches. Close to probabilistic modeling, we have also the neural network method. In this paper, we focus on the application of neural network modeling for seismic pattern recognition (seismic facies analysis) applied an ultra-deepwater turbidite oilfield reservoir in Campos Basin, offshore Brazil. Introduction A 3-D reservoir architecture characterization requires the integration of different data types to define a more detailed and realistic geological interpretation. Well logs and core data provided detailed information about the vertical variation of many reservoirs properties but they are restricted to regions adjacent to the borehole. 3-D seismic data play an important role in describing external and internal complexities of reservoirs away from the wellbore and to define geometric description of structural and stratigraphic aspects of the reservoirs (ref. 1). Seismic amplitude variations are linked to changes in acoustic impedances that we can be trying to relate to reservoir properties. This paper demonstrates a methodology for seismic pattern recognition in a targetoriented approach to aid the reservoir architecture characterization in a more detailed and accurate 3-D seismic interpretation. Seismic pattern recognition techniques are used to distinguish important geological features from seismic information. The methods of seismic pattern recognition can provide solution to practical problems in reservoir characterization in terms of automatic mapping of main features of seismic morphology related to geological environment. The automatic interpretation of subsurface geology from seismic data is possible by analyzing of the nature of waveform cycles and their termination with respect to adjacent reflections. The geometry and the terminations of waveform styles help to locate boundaries between zones corresponding to different types of depositional units each associated with characteristics of seismic morphology under study. In this paper a target-oriented automatic pattern recognition methodology is applied to 3-D seismic data set a seismic stratigraphic unit of an ultra-deepwater turbidite sandstones reservoir. The pattern recognition method is applied in two approaches: unsupervised and supervised. The unsupervised approach to exploit the statistically common characteristic underlying seismic traces segments at the seismic stratigraphic unit. The supervised approach uses the stratigraphic knowledge to guide the pattern recognition. The seismic pattern recognition methodology used is carried out in six main steps: (1) spatial and temporal sampling, (2) attributes selection; (3) definition of the number of classes and iteration; (4) training and classification with a competitive learning algorithm (unsupervised approach) and (5) training and classification with a back-propagation algorithm (supervised approach) and (6) interpretation of seismic facies. Work on artificial neural network has been motivated from the results obtained in terms of useful computation of learning process of seismic waveform. To achieve good performance, neural network employ a massive interconnection of simple computing cells referred to as neurons or processing units. The SPE 69483 Reservoir Geophysics: Seismic Pattern Recognition Applied to Ultra-Deepwater Oilfield in Campos Basin, Offshore Brazil Paulo Johann, Dayse D. de Castro and Alberto S. Barroso, Petrobras S. A. 2 JOHANN, P.R.S.; CASTRO, D. AND BARROSO, A. SPE 69483 definition of neural network is a massively parallel distributed processor that has a natural propensity for storing experimental knowledge and making it available for use (ref. 2). Geology data set description The ultra-deepwater turbidite oilfield analyzed in this study was discovered in October 1996, located 125 miles from the coast in the northeast portion of Campos basin, offshore Brazil, in water considered to be ultra-deep (between 1,5002,000m) (Fig. 1). Campos basin is located in the southeastern coast of Brazil and it covers an area of about 100,000 km from the coast to the 3,400m isobath. The Vitória high separates it from Espírito Santo basin to the north and from Santos basin by the Cabo Frio high to the south. This field is a large oilfield with a complex hydrocarbon distribution (OOIP around 2.0 billions m with API from 18.6o to 31.5o). The external geometry of the field is defined to the north and east by dipping and to the south and west directions by stratigraphic pinchout. The oil entrapment is composed by structural and stratigraphic framework. Two main factors controlled deep and ultra-deepwater sedimentation in Campos basin: thermal subsidence pattern which control turbidite sedimentation to certain preferable areas and salt movement which allowed stacking of sandstones in depositional lows. Unconformities due to sealevel variation and submarine paleocanyons are the additional factors that control reservoir distribution (ref. 3). Hydrocarbon accumulation are Miocene to Maastrichtian ages. The lithology is a turbidite sandstones deposited in a widespread structure depression. The halokinesis process controlled the structural framework. The stratigraphic zonation from well logs measurements divided the reservoir in three main sequences, with internal sub-division (ref. 4). Five main zones can be characterized with average thickness of 30m by zone. Total oil net sand in the field is very thick, with average around 160m. In this paper we focus in the Maastrichtian 1 reservoir zone. This reservoir was used in the beginning of the delimitation of the field to build the initial structural map and to estimate from seismic data the reservoir distribution (ref. 5). Development strategy overview The decision by Petrobras to develop the field immediately after its discovery was based upon the success of the drilling of exploration wells (ref. 3). The improved of 3-D seismic data and the technological training program that has given Petrobras the capability to produce oil and gas in ultradeepwater in Campos basin. The Pilot System start on January 1999 and the field became the holder of world record for oil production in ultradeepwater (1,853m water depth). The objective of Pilot System was to collect fundamental information about the reservoir and test new technologies to be applied in the production system. Also this system has furnished information that allow optimization of the subsequent field exploitation stages, thus helping reduce technical, economic and environmental risks during these phases, when large volumes of oil and gas are being produced. The Phase1 of permanent production system start on April 2000 and produce the North and Southeast portions of the reservoir. Between the first 9 wells drilled in this phase of production, that confirmed the oil with excellent quality 31.5o API, Petrobras had the world record for oil production in ultradeepwater (1,877m, Fig. 1). Principles of methodology seismic facies analysis The pattern recognition methodology is carried out in six main steps (ref. 7): (1) spatial and temporal sampling, (2) attributes selection; (3) definition of the number of classes and iterations; (4) training and classification with a competitive learning algorithm (unsupervised approach) and (5) training and classification with a back-propagation algorithm (supervised approach) and (6) interpretation of seismic facies. Spatial and temporal sampling of the data set. The 3-D seismic volume available over the reservoir has 414 km. The cell dimension is 13,6m by 26,6m, with a spatial density of 100.000 traces/km. The record length is from 0 to 6 seconds, with 2ms of sample rate. The seismic data used in the pattern recognition is migrated pre-stack in time. The first step in the methodology is the choice of the representative sampling of the seismic data available over the reservoir. The reservoir external geometry is the first point to take in account to define the area of analyzes. In our study the external geometry of the Maastrichtian 1 reservoir was the guide to define the polygon of study. Figure 2 shows the spatial volumetric distribution of Maastrichtian 1 reservoir in terms of seismic amplitude prestack migrated in time. The focus in the area of the reservoir distribution reduces to 130 km (367.460 seismic traces). The data to apply the pattern recognition algorithm. The temporal sampling was a sub-volume cut from 10ms above and 30ms below the Maastrichtian 1 reservoir top, respectively. The window of 40ms used in the pattern recognition also reduces the input data from 3000 samples/trace in the raw data to 20 samples/traces under study (7.349.200 seismic samples). The seismic horizon representative of Maastrichtian 1 reservoir was carefully picking before the definition of the temporal window for the pattern recognition. Figure 3 shows the temporal window over a representative seismic line used for pattern recognition algorithm. Attributes selection. Six volumetric seismic attributes were selected for carried out the seismic facies analysis. All attributes were calculated over the volume inside the window around t

Seismic interpretation of self-organizing maps using 2D color displays

Classification without supervision of patterns into groups is formally called clustering. Depending on the application area these patterns are called data lists, observations or vectors. For exploration geophysicists, these patterns are usually associated with seismic attributes, seismic waveforms or seismic facies. The main objective of this paper is to show how one of the most popular clustering algorithms - Kohonen self-organizing maps, can be applied to enhance seismic interpretation analysis associated with one and two-dimensional colormaps.

Spectral decomposition reveals geological hidden features in the amplitude maps from a deep water reservoir in the Campos Basin

In Petrobras’s research project “New Technologies Applied to Reservoir Characterization of Thin Turbidites Reservoirs”, PRAVAP 19 Advanced Oil Recovery Program, there is a special interest in the use of this technology in order to obtain an estimation of the thickness especially in the presence of stratigraphic pinch-outs that may be causing tuning effect in some reservoirs. These geological situations represent important oil volumes in the main reservoirs of the company.

Brazilian Deep Water Carbonate Reservoir Study Using the Wavelet Transform Teager-Kaiser Energy

Spectral decomposition has proven to be a powerful means to identify strong amplitude anomalies at specific frequencies that are otherwise buried in broad-band response. We compute Teager-Kaiser Energy for each component of a joint time-frequency representation to generated from a 3D survey acquired over a Brazilian deep water carbonate reservoir. This nonlinear energy tracking algorithm allows us to differentiate between high amplitude reservoir and other high amplitude reflections. We calibrate our algorithm against synthetic seismic traces generated from the well logs and and then apply to the real seismic data to reveal important geological features.

Ultra-deepwater 4-C offshore Brazil

This paper is the case history of an ultra-deepwater 4-C seismic program, recently acquired in the Campos and Santos basins, offshore Brazil. All aspects of the program will be discussed, including the feasibility study which first indicated the potential value of multicomponent technology to this region, to the survey objectives, to the illumination analysis used in designing the survey, to the field operations, to the fully processed results. Acquisition was completed on 19 April 2005, and the fully processed data were delivered in early November 2005.

Relative acoustic impedance from wavelet transform

Most deconvolution algorithms try to transform the seismic wavelet into spikes by designing inverse filters that remove an estimated seismic wavelet from seismic data. We assume that seismic trace subtle discontinuities are associated with acoustic impedance contrasts and can be characterized by wavelet transform spectral ridges, also called modulus maxima lines (WTMML), allowing us to improve seismic resolution by using the wavelet transform. Specifically, we apply the complex Morlet continuous wavelet transform (CWT) to each seismic trace and compute the WTMMLs. Then, we reconstruct the seismic trace with the inverse continuous wavelet transform from the computed WTMMLs with a broader band complex Morlet wavelet than that used in the forward CWT. Because the reconstruction process preserves amplitude and phase along different scales, or frequencies, the result looks like a deconvolution method. Considering this high-resolution seismic representation as a reflectivity approximation, we estimate the relative acoustic impedance (RAI) by filtering and trace integrating it. Conventional deconvolution algorithms assume the seismic wavelet to be stochastic, while the CWT is implicitly time varying such that it can be applied to both depth and time-domain data. Using synthetic and real seismic data, we evaluated the effectiveness of the methodology on detecting seismic events associated with acoustic impedance changes. In the real data examples, time and in-depth RAI results, show good correlation with real P-impedance band-pass data computed using more rigorous commercial inversion software packages that require well logs and low-frequency velocity model information.