A multimethod Global Sensitivity Analysis to aid the calibration of geomechanical models via time-lapse seismic data

Abstract

Time-lapse seismic attributes are used extensively in the history matching of production simulator models. However, although proven to contain information regarding production induced stress change, it is typically only loosely (i.e. qualitatively) used to calibrate geomechanical models. In this study we conduct a multimethod Global Sensitivity Analysis (GSA) to assess the feasibility and aid the quantitative calibration of geomechanical models via near-offset time-lapse seismic data. Specifically, the calibration of mechanical properties of the overburden. Via the GSA, we analyse the near-offset overburden seismic traveltimes from over 4000 perturbations of a Finite Element (FE) geomechanical model of a typical High Pressure High Temperature (HPHT) reservoir in the North Sea. We find that, out of an initially large set of material properties, the near-offset overburden traveltimes are primarily affected by Young's modulus and the effective stress (i.e. Biot) coefficient. The unexpected significance of the Biot coefficient highlights the importance of modelling fluid flow and pore pressure outside of the reservoir. The FE model is complex and highly nonlinear. Multiple combinations of model parameters can yield equally possible model realizations. Consequently, numerical calibration via a large number of random model perturbations is unfeasible. However, the significant differences in traveltime results suggest that more sophisticated calibration methods could potentially be feasible for finding numerous suitable solutions. The results of the time-varying GSA demonstrate how acquiring multiple vintages of time-lapse seismic data can be advantageous. However, they also suggest that significant overburden near-offset seismic time-shifts, useful for model calibration, may take up to 3 yrs after the start of production to manifest. Due to the nonlinearity of the model behaviour, similar uncertainty in the reservoir mechanical properties appears to influence overburden traveltime to a much greater extent. Therefore, reservoir properties must be known to a suitable degree of accuracy before the calibration of the overburden can be considered.

Electrokinetic Contributions to Self‐Potential Signals From Magmatic Stressing

“…Here Y is the associated 1 × j output matrix for each model run, while x max and x min define the maximal and minimal range of each input factor, respectively. This calculation is repeated for n model runs to obtain n EET (EET n,i ) for each input parameters (Price et al, 2018). The EET sensitivity can be assessed by the mean, which denotes the degree of influence of an input factor, and the standard deviation, evaluating the interaction between parameters.…”

Time‐lapse VSP integration and calibration of subsurface stress field utilizing machine learning approaches: A case study of the morrow B formation, FWU

“…Several publications on the calibration of a coupled hydro-mechanical model using attributes computed from a VSP dataset have been found. 13,[19][20][21] Attributes such as the seismic wave travel times, velocities, amplitudes, and seismic reflectivity have been utilized to update reservoir flow properties and monitor the preferential fluid movement. 4,5,11,22 For instance, In Salah, a well-known CCUS project in Algeria has CO 2 stripped from produced natural gas and reinjected into the Krebka formations' aquifer leg.…”

“…Several publications on the calibration of a coupled hydro‐mechanical model using attributes computed from a VSP dataset have been found 13,19–21 . Attributes such as the seismic wave travel times, velocities, amplitudes, and seismic reflectivity have been utilized to update reservoir flow properties and monitor the preferential fluid movement 4,5,11,22 .…”