Sammendrag
There are relatively few large-scale projects when it comes to CO2 storage, which means that it is necessary to increase knowledge about CO2 flow, as well as validate and calibrate the models and workflows we use. This is where FluidFlower comes in as a tool where, unlike a real reservoir, we have control over model parameters and the opportunity to visually observe the CO2 dynamics in the reservoir. Access to high-resolution images of CO2 flooding provides a unique opportunity to validate the simulation models as well as explore parameter sensitivity and workflows for history matching. In this talk, we present results of performing history matching studies on the benchmark FluidFlower data set using OPM Flow for the numerical simulations and ERT as the history matching toolbox. First, we introduce the open workflow `pyff’, a Python package to set up the studies via a configuration file. This gives researchers and lecturers a simple and transparent way to compare experiments and modeling. Second, we examine different setups for history matching of tracer and CO2 data. In particular we compare using pixel based data at some given time-intervals with using averaged quantities in some given boxes. The pixel based data is motivated from seismic while the average box values resembles well type data.
https://meetings.siam.org/sess/dsp_talk.cfm?p=129527
Vis fullstendig beskrivelse