Integrating Time-Lapse (4-D) Seismic Data with Reservoir Simulation" by Dr. John Waggoner

Abstract
Reservoir simulation is one of the most used reservoir management tools, but it can be no better than the reservoir model given to it as input. However, time-lapse 3D, or 4D, seismic data is a spatially resolved measurement that can give an indication of fluid movement in the reservoir. By integrating the 4D seismic results with reservoir simulation, it is possible to create a reservoir model that matches both the measured static and dynamic properties of the reservoir. The intent of this talk, then, is to describe some of the ways that 3D seismic data, and specifically time-lapse 3D, or 4D, seismic data, are being used to improve reservoir models.

After briefly presenting the fundamentals of 4D seismic, from the perspective of a reservoir engineer, this talk will focus on some of the ways of integrating 4D seismic data with reservoir simulation. Case histories will illustrate the concepts and methods presented, ranging from visual interpretation to model screening to seismic history matching. No one method is appropriate for all cases, so the strengths and limitations of each will be discussed.

Biography

John Waggoner is Reservoir Engineering Manager in the Reservoir Services Group at Western Geophysical in London. He holds BS, MS, and PhD degrees in Petroleum Engineering from The University of Texas at Austin, and worked 7 years at Sandia National Laboratories in the Geophysical Technology Department before joining Western Geophysical in 1997. His work experience includes reservoir characterization, artificial lift, and borehole telemetry. His current work focuses on integrating time-lapse (4D) seismic data with reservoir simulation, and on the use of 3D seismic data for improved reservoir characterization.

Time-lapse 3D, or 4D, seismic is gaining acceptance as a valuable element of effective reservoir management.  When it works, 4D seismic can reveal undrained reservoir compartments, identify flood front positions, and assist the evaluation of potential drilling targets.  These are examples of what this presentation refers to as Visual Interpretations of 4D results.  Assuming that a 4D result has been obtained (see, for example, Waggoner, 1998), this presentation focuses on the different ways that the 4D result can be used, and in fact, integrated, with reservoir simulation.  Each method discussed will be illustrated during the presentation by recently published case studies.  However, due to the difficulty in reproducing detailed color images, only a reference will be made to those published studies where the interested reader can find both more information about the study and clear graphical images.

Methods of using 4D results

There are 4 primary ways in which 4D results are being used:

1.      Visual Interpretation

2.      Quantitative Interpretation

3.      Model Screening

4.      Seismic History Matching

Visual Interpretation

By far the most common use of 4D results in the literature to date has been visual interpretation.  Simply stated, visual interpretation presents the 4D result as volumes or sections of difference data.  Observations are checked with well and production information to gain confidence in the meaning of the different anomalies.  In addition, the volumetric continuity of anomalies is checked as a way of potentially excluding difference noise.  At the end of such an investigation and validation phase, the 4D results can be interpreted as production changes within the reservoir.

Two recent case studies illustrate the value of visual interpretation.  The first (Koster, et. al., 2000) is the highly successful Gannet-C study in the UK North Sea Central Graben.  The reservoir consists of a ring shaped closure around a central salt dome, and consists of numerous radial faults with uncertain sealing.  Pressure is maintained by natural water drive and a single gas injector, so observed changes can be interpreted as saturation effects and not pressure effects.  The rock physics model predicts a 25% changed in the top reservoir amplitude as a result of the saturation change.

When the difference amplitude is superimposed on the top reservoir depth map (Fig 7 in Koster, et. al., 2000), drained areas of the reservoir are clearly indicated.  Also clearly seen is the lack of change in a compartment bounded by radial faults, and the presence of change in an area outside the original oil water contact initially thought to be in the water leg.  The management of the Gannet-C reservoir has benefited greatly from this information, leading to a drilling program to access the estimated 15 million bbl of oil in the undrained compartment.

The second case study (Kristensen and Seymour, 1998) is the giant Statfjord field study straddling the Norwegian and UK sectors of the North Sea in the northern part of the Viking Graben.  Water injection into the aquifer has helped to maintain pressure, but both pressure and saturation changes are expected to change the acoustic impedance in the reservoir by up to 10%.  Among other observations in this study, a drilling target location was shown to be in a drained zone, and thus was not drilled.

Quantitative Interpretation

Quantitative interpretation involves the conversion of the calculated seismic change into a reservoir change.  In one approach (Kvamme et. al., 2000), the calculated change in reflection strength is calibrated at the well locations to changes in water saturation.  A remarkably strong correlation (coef. = -0.79) indicated that such an approach was feasible, so a water saturation change map was generated using the geostatistical collocated cokriging technique, guided by the change in reflection strength.  The water saturation changes were then subtracted from the absolute water saturation values at the base survey time to yield the water saturation map at the monitor survey time.  Five new wells were drilled after the monitor survey time, and the water saturation values predicted from 4D were closer to the observed water saturation values than were the water saturation values predicted by the Eclipse simulation model.

Model Screening

Model screening refers to comparing, in some sense, the measured seismic change against predictions from multiple reservoir models. The multiple models may come from multiple geostatistical simulations of reservoir properties, or they may come from different geologic interpretations. In either case, the models are considered to be equally likely, and the data available cannot rule out any of them.  It is the objective of model screening to provide additional information to eliminate some of the models from further consideration.

The comparison between modeled and measured 4D effect can take many forms.  In the Draugen study (Koster, et. al., 2000), the comparison was made on the change in equivalent hydrocarbon column (DEHC), which is essentially a measure of saturation change. DEHC is obtained from the measured seismic through a proprietary Shell inversion technique. DEHC is predicted from the multiple models by simulating each one and calculating the resulting DEHC between the base and monitor seismic survey times. 

In the Draugen field, water injectors at the north and south end of a low relief anticline structure pushed oil toward the central producers.  The lack of wells between the producers and the injectors resulted in a lot of uncertainty in reservoir properties, leading to at least 3 different reservoir models.  Each model shows water coming up both the east and west flanks of the reservoir from the southern injectors, and each modeled a northern fault as a transmissive fault.  The measured 4D result, however, clearly shows the injected water moving along the western flank, with no displacement on the eastern flank.  Also, the northern fault is clearly shown to be sealing over production time.  While all models were in error, the model that was closest to the 4D result was selected for further study, including some seismic history matching, as will be discussed next.

Seismic History Matching

Seismic history matching is directly analogous to conventional history matching. However, instead of comparing only the measured and simulated production data, the difference between the measured and synthetic 4D data are also considered.  In the South Timbalier study (Huang, et. al., 1997), the simulation results at the times of the base and monitor seismic surveys were forward modeled into a predicted seismic change by first applying a Gassmann fluid substitution to get acoustic impedance, and then a zero-offset convolution model to get amplitude.  This synthetic 4D amplitude result is then compared with the measured 4D amplitude result to help determine where the reservoir model should be modified to achieve a better history match.

As with production data history matching, seismic history matching can be automated to further assist the history matching process.  In this case, the difference between the measured and synthetic 4D data is quantified as the correlation between the 2 data sets, much as the difference in production data is quantified as the sum of least squares differences.  Both the production and 4D quantified differences can then be combined into a single objective function for use in an optimization algorithm.

Summary

Figure 1 provides a means of summarizing the different approaches described in this presentation.

Visual interpretation begins with the Seismic Data and moves down to Amplitude or Acoustic Impedance (or some other attribute).  The resulting images are interpreted for changes and production induced effects.

Quantitative interpretation goes through the next inversion step to water saturation (or some other reservoir property).  The resulting absolute and delta water saturation maps are used to indicate what areas have been swept, but also to quantify how much oil might remain.

Model screening starts with multiple Reservoir Models and simulates each to predict changes that correspond to the measured seismic change.  For example, if a quantitative interpretation has produced a water saturation change map, then the reservoir models can be screened based on the water saturation maps produced by simulation.  However, the simulator results can also be taken through a forward modeling sequence to Acoustic Impedance or Amplitude, and screened against the seismic results at earlier stages of processing.

Seismic history matching is similar to the description of Model screening above, except that a single Reservoir Model is considered and updated as a result of the comparisons made.  Again, comparison between predicted and measured data can be made at Amplitude, Acoustic Impedance, or water saturation levels, depending on what gives the best 4D difference image to work with.  Production Data is shown because seismic history matching can match both the 4D and production simultaneously by combining both elements in the objective function.

Conclusions

1.      Time-lapse seismic can impact reservoir management decisions.

2.      Time-lapse seismic data can be used with production data to improve the reservoir model.

3.      4-D is, at present, the only viable means of constraining the reservoir model between & beyond the wells.

Figure 1:  Flowchart summarizing the various uses of 4D seismic data.

References

Huang, X., Meister, L., and Workman, R.: “Reservoir Characterization by Integration of Time-Lapse Seismic and Production Data,” SPE 38695 presented at the 1997 SPE Annual Technical Conference and Exhibition, San Antonio, TX, 5-8 October, 1997.

Koster, K., Gabriels, P., Hartung, M., Verbeek, J., Deinum, G., and Staples, R.: “Time-Lapse Seismic Surveys in the North Sea and Their Business Impact,” The Leading Edge, March 2000.

Kristensen, A. and Seymour, B.: “Statfjord Field 4D Reservoir Monitoring: Project Plan and Initial Results,” presented at PETEX98, London, UK, 1-3 December 1998.

Kvamme, L., Al-Najjar, N., Psaila, D., Astratti, D., and Doyen, P.: “Statfjord Field Saturation Mapping from Time-Lapse Seismic Data,” presented at NPF, Kristiansand, Norway, 13-15 March 2000.

Waggoner, J.: “Lessons Learned from 4D Projects,” SPE 49144 presented at the 1998 SPE Annual Technical Conference and Exhibition, New Orleans, LA, 27-30 September 1998.