Fully coupled inversion on a multi-physical reservoir model - Part I: Theory and concept

State of the art reservoir monitoring delivers numerous property data with high resolution. Especially the consistent interpretation of pressure data with different geophysical methods requires multi-physical modelling and inversion workflows. Such a workflow is developed based on the reservoir monitoring concept of the Ketzin pilot site for CO2 storage, Germany. The workflow consists of three physical models, (i) a single phase hydraulic model, (ii) a multiphase CO2 migration model and (iii) a geoelectrical model. Calibration is carried out to match observation data groups hydraulic pressure, CO2 pressure, CO2 arrival time and geoelectrical cross-hole observations. Calibration parameters are spatially distributed hydraulic permeability and porosity, compressibility, the relative permeability function and the geoelectrical saturation exponent. Geoelectrical measurements with low coverage that cannot be inverted with traditional methods could be included, since the multiphysical reservoir model acts as physical regularisation. The indirect nature of geophysical data is overcome by implementation of petrophysical relations between permeability and porosity and between CO2 saturation and electrical resistivity. Stability against field data is increased by reducing the impact of structural noise through preprocessing the observation data. Stability against the overparameterisation is added by Tikhonov regularisation and singular value decomposition, the latter combined with super parameter definition reducing the problem dimensions and simulation time by three quarters. A synthetic case study demonstrates that the model resolves the spatial permeability and identifies the petrophysical relation between CO2 saturation and electrical resistivity. The weighting scheme balances different observation data groups and measurement intervals. The model to measurement misfit is reduced proprotionally for all observation data groups, while the geoelectrical data are most difficult to match.