Using full wave seismic modeling to test 4D repeatability for Libra pre-salt field

Abstract

In the scope of the development of the MERO (Libra NW) project, Libra Joint Project Team and Petrobras are preparing a project of 4D acquisition on MERO based on a permanent reservoir monitoring (PRM) network to be installed at the seabed in early production phase (less than 2 years after first oil). With more than 6500 4C stations planned, such a pre-salt PRM is a novel and ambitious project, the first of its kind in a pre-salt setting. Pre-salt 4D signal is weak and 4D repeatability noise in particular can be affected by complex salt propagation effects. Considering these challenges, the costs and reservoir stakes of the project, an extensive full wave acoustic and elastic 4D seismic modeling program has been designed to study geometric repeatability issues as well as prepare future estimation of gas-sensitive 4D attributes and integrate 4D workflow. Preliminary results are discussed Introduction Libra pre-salt field, located in Santos basin, contains large resources to be produced over more than two decades, over a relatively compact area a priori favorable to a PRM setup. Pre-salt reservoirs contains distinct carbonate rocks – stromatolites, grainstones and coquinas, all from Aptian age. Reservoirs are thick yet interspersed with igneous rock. Overall, the reservoir units exhibit four orders of magnitude of permeability variations. A key challenge is the very high GOR in Libra NW, with a CO2-rich gas that poses a great challenge for exporting the HC gas due to restricted economic conditions. A first Extended Well Test already proved excellent pressure communication but its matching reinforced the role of potentially extended multi-darcyan layers, a potential pathways for early gas breakthroughs and a challenge for gas management. Gas capacity will therefore be the main bottleneck to production and efficient gas management key to project NPV. The motivations for investing in a frequenct high density/high repeatability 4D scheme on Libra, stems from the anticipated added value of 4D information to directly assist a production drive, based on WAG alternating water and gas reinjection perhaps combined with other mechanisms. An important aspect of Libra reservoir management is the flexibility of drive obtained using Intelligent Completions Valves (ICVs) with two to three independent levels on injectors and producers. On the 4D side, this implies that on top of the conventional added value of 4D associated to second wave or infill drilling, the vision of fluid fronts via high quality 4D seismic would also be a key enabler to a more efficient reservoir management via better reservoir understanding and more proactive use of ICVs, with the main objective of preventing or delaying early gas breakthroughs. But one needs to see gas fronts. 4D signal is not bound by 3D resolution that is solely dependent on useful signal bandwidth: it is equally modulated by 4D noise. It can detect very subtle dynamic features, on condition of being above 4D noise, that is dominated by repeatability noise (2): Improving repeatability is the way to go. Yet combining weak signal in thin streaks with the complexity of seismic imaging of pre-salt reservoirs builds a real challenge for 4D detection, a challenge reinforced by the likely necessity to extract 4D AVO attributes such as Ip Is (Acoustic Shear impedance) or dPr/Pr (Poisson Ratio) to maximize fluid sensitivity and visualize, as shown below, thin gas streaks despite low fluid contrast. Figure 1 – dIp/Ip 4D random line across WAG wells and same with dPr/Pr 4D Poisson ratio attribute, from PEM. SHORT TITLE (50 LETTERS MAXIMUM. FONT: ARIAL 9) ____________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Sixteenth International Congress of the Brazilian Geophysical Society 2 High repeatability such as demonstrated by PRM experience of various operators in the post salt being necessary, a PRM is the anticipated answer to Libra 4D challenge on reception side. But will it be sufficient ? And beyond that, what type of workflow should we use to extract the most usable 4D AVO signal idealized in Fig. 1, with Libra ́s complex top salt topography having a profound impact on wave propagation and imaging? To anticipate answers to these difficult questions and provide credible elements of appreciation to Libra asset before making a final investment, we have launched an innovative 4D full wave acoustic and elastic modeling study, incorporating as much realism as practical as of today. The modeling phase of the study has only been running for a few weeks at the time of writing and this expanded abstract will primarily focus on the methodology of model building and modeling, on the setup for testing of various differential acquisition geometries and on calibration and comparisons to real data. However, we will integrate more elements of full wave 4D modeling results as per their availability. Model building Typical 4D feasibilities combine reservoir simulation based on production scenarios with petroelastic modeling to find relevant 4D observations to address production challenges, as illustrated in Figure 1 for Libra. Following this line a second level study (1) was based on the incorporation of 4D noise according to a noise model and methodology developed by Total (2).It showed that although weak, usable 4D signal could be obtained on Mero, assuming a NRMS level around 2.5 %, a pre-salt, a lowfrequency equivalent value, as proposed by PGS(3) to the 3-4 % that has already been obtained by Petrobras with pilot PRM setup in Jubarte post-salt field (4). However, moving from post-salt to pre-salt, seismic propagation becomes as complex as the salt topography implies. In such a setting the 1D convolutional approach appears as a very crude approximation. Firstly, 1D convolution does not enable to easily integrate important acquisition limitations such as actual source and received density and geometry. Those impact 3D illumination and potentially affect 4D sensitivity to acquisition geometry misfits. Such effects can be modeled accurately by simulating base and monitor acquisition with 3D full wave acoustic or elastic modeling. Secondly, the 1D convolutional approach does not apprehend the impact of imaging on local blurring, mislocation and lack of amplitude fidelity of the 3D or 4D response, but by actually simulating seismic shots and performing reverse time migration (RTM) with a suitable velocity field, one reproduces these aspects in the simulation results. To validate the design, it was therefore decided to follow a more ambitious approach, similarly to a paper published in 2016 by Total on a subsalt target (5). Therefore an area of 24.5 x 24 km, providing good imaging conditions for the widest part of Libra NW (Mero Field) core area, was defined for building a model for full wave seismic modeling. The vertical extension of the model covers from sea surface to a depth of 7 km, deemed sufficient given the quality of the boundary conditions that can be applied to full wave acoustic and elastic propagation. Figure 2 – Model area showing (left) base salt and (right) top salt iso-depth surfaces. The workflow for model construction has been designed using available well data and the following volumes from Libra ́s 2016 reprocessing of the streamer SPEC survey: FWI velocity volume and tomography volumes of Epsilon and Delta. 35 Hz RTM stack (used from top salt downwards) 3 m Kirchoff PSDM (used above top salt) Intercept and Gradient extracted form the Kirchoff (used above top salt). AUTHORS (50 LETTERS MAXIMUM. FONT: ARIAL 9) ____________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________________ Sixteenth International Congress of the Brazilian Geophysical Society 3 The modeling workflow is illustrated below. Figure 3 – Schematic workflow for elastic model building An hybrid "KIRTM" dataset has first been generated by merging the two stack volumes delimited by top salt as a key surface. It was then corrected for illumination using a first-order volumetric correction consisting of a linear interpolation from constant seabed illumination to variable top salt, then to base salt illumination. Top salt illumination is important to obtain more faithful properties from amplitude. At base salt it was mainly implemented as a first order correction (map-based and not volumetric) to avoid introducing NAZ biases between the petroelastic description of the reservoir properties and the high frequency part of the seismic-driven elastic properties. The KIRTM stack data has been used to derive the high-frequency part of an impedance model using the Iterdec © pseudo-impedance generation. This high frequency part has then been used as a modulation to a low frequency model based on FWI Vp velocities and, for each major lithology interval, Gardnerlike relations relating Rho to Vp and Vs to Vp defined using a set of well logs. The image below compares a Vp section between low frequency, impedance-modulated and AVOmodulated Vp. Figure 4: Evolution of Vp volume from FWI, integrating impedance and AVO contributions Using this approach, a full band acoustic impedance model was first built, then it was mapped from impedance to Vp, Vs and density full band cubes using direct mapping from A