Quantitative evaluation of structural compartmentalization in the Heidrun field using time-lapse seismic data

Abstract

In reservoir settings with structural compartmentalization, fault properties can constrain the fluid flow and pressure development, thus affecting decisions associated with the selection of the drainage strategy within reservoir management activities. Historically, we have relied on geological analysis to evaluate the fault seal, however this can be restricted by available well coverage which can introduce considerable uncertainty. More recently, time-lapse seismic has become useful in the assessment of the dynamic connectivity. Indeed, seismic changes are in general a combination of pressure and saturation changes which, for compartmentalized reservoirs, seem to be associated with the sealing behaviour of faults. Based on this observation, this thesis presents a new effort in which the spatial coverage of the time-lapse seismic data is used as an advantage to more fully resolve properties of the fault seal, particularly in areas with poor data control. To achieve this task, statistics of amplitude contrast and the spatial variability of the 4D seismic signatures are considered. Tests performed on modelled data have revealed that the proposed 4D seismic measurements can be calibrated at the wells in a sector with known geological characteristics via a quadratic polynomial expression that allows fault permeability to be derived. Uncertainties in the 4D seismic estimation have also been considered in a Bayesian framework, leading to the identification of error bounds for the estimates. Results on synthetic data are encouraging enough to investigate its applicability on the Heidrun field. In this real example, the Jurassic reservoirs are compartmentalized due to the presence of a set of faults for which their flow capacity strongly affects field depletion. Here, previous studies have attempted to characterize the fault seals, yet the sparse nature of well data has limited their evaluation, leaving uncertainties when adjusting fault properties in the reservoir simulation model. In this case, application of our approach has proven useful, as it has allowed the detailed characterization of major faults in this field. Predictions obtained with the 4D seismic appear consistent when compared to previous core observations made from fault-rocks studies. Also, the results have been used to update ii the flow simulation model by adjusting transmissibility factors between compartments, leading to a decrease of the mismatch between the simulated forecast and historical production data. Furthermore, uncertainty in the 4D seismic prediction has been considered when implementing an automatic history match workflow allowing further improvements. New insights into the implications of the dynamic fault behaviour in the time-lapse seismic response are also provided in this thesis. We make use of synthetic models in which faults represent the main constraint for fluid flow, to show that an adjustment of the relation between the reservoir capillary pressure and the capillary threshold pressure of the fault-rock can alter the variance of the time-lapse seismic signature. However, a similar behaviour can be obtained when strong variations in the transmissibility of the fault are present. As a consequence, we propose that this statistic might help to identify fault seal dependent controls on individual fluid phases when the transmissibilities are fairly similar along the fault segment. This is particularly useful in the Heidrun field where we have found it difficult to explain the water encroachment by only using the single-phase approximation offered by the fault transmissibility multipliers. Here, the variance of the 4D seismic signature is employed together with the fault permeability values to suggest that in some compartments, waterflooding might be affected by the presence of a specific fault with sealing capacity strongly dependent on the individual fluid phases. This helps to explain the observed fluid uncertainties. It is also recognized that more data might be required to gain greater insight into this issue; hence alternative hypotheses are not discarded.