Dynamic simulation and history matching at Ketzin (CO2SINK)

Abstract Since the end of June 2008, carbon dioxide (CO 2 ) is being injected in the Stuttgart Formation at Ketzin, Germany as part o f the European Union’s CO2SINK project. The injection well (Ktzi 201) is roughly 50 and 100 m away from observation wells Ktzi 200 and Ktzi 202, respectively. CO 2 was detected at the closest observation well on the 15th of July 2008, approximately 20 days after injection commenced. Breakthrough at the Ktzi 202 well was recorded on March 21, more than 8 months after injection was initiated. Dynamic simulations of the injection and flow of CO 2 into the subsurface at Ketzin will be described. The three dimensional (3D) geological model was built based on a 3D seismic survey as well as logging and core analysis data. The formation consists of fluvial sandstone channels within a muddy flood plain at mean depth of 650 m within an anticlinal structure. The geological model uses a cell size of 20×20 m and a layer thickness of 0.5 m, resulting in a total of ∼7 million cells. The model is upscaled (i.e., coarsened) and the grid is refined locally in the zone comprising the injection and observation wells. This upscaled model has corner point geometry with 78×74×91 (525, 252) grid cells in X, Y, and Z directions. X and Y direction grid sizes around the wells are respectively 4 m and 5 m. Average grid size in Z direction is about 0.5 m. Following the upscaling of the geological or static model, a blackoil commercial streamline simulator was used to simulate the flow of injected CO 2 according to the actual injection rate history. The phase behavior of CO 2 and brine were described by blackoil pressure-volume-temperature (PVT) tables. The blackoil PVT was modeled by imposing brine properties to the simulator oil model and CO 2 properties to the simulator gas model. In this manner, solubility of CO 2 into water was taken into account. Salinity of the brine is represented by the appropriate density. Fluid properties (density, viscosity) are pressure dependent and isothermal (at reservoir temperature). The outcome of the simulation is the time for injected CO 2 to arrive at the observation wells and the history of bottom-hole pressures (BHP) as a function o f time that can be compared to measurements taken in the bottom of the injection well. History matching was performed by adjusting the permeability (along and across bedding) until model BHPs agreed with the measured ones. Permeability was chosen as history matching parameter because of its high degree of uncertainty and relies on the porosity/permeability relation derived from core measurements. Good agreement was obtained with a multiplicator of 0.1 applied to the permeability across and along bedding. Breakthrough time for the closest observation well was in good agreement with reality. However the breakthrough time at the most distant well was underestimated by several months. We believe that geological features at distances greater than 50 m from the injection well may be responsible for the mismatch and should be investigated further.