| Title | Joint Kriging of geothermal reservoir properties -methods and challenges |
|---|---|
| Authors | Gu, Rühaak, Bär, Sass |
| Year | 2013 |
| Conference | European Geothermal Conference |
| Keywords | thermal conductivity, kriging with external drift, geothermal reservoir, univariate/bivariate spatial description, geostatistic modelling |
| Abstract | Estimating petrophysical properties of geothermal reservoirs is always a challenging task. The highest interest is typically on permeability and thermal conductivity, which are especially important parameters for numerical reservoir models. Typically only very few data exist while the budget is often small. To improve the understanding an extended and joint data exploration is therefore of vital importance. In a pilot study the potential of extended data analysis methods is demonstrated for a Rotliegend reservoir in Germany. Along a 47 m long core measurements with a distance of only 1 mm were performed, summing up to 32,000 data points. This dense data distribution allows for studying the spatial continuity and variability in detail using one-dimensional semivariograms (spatial variance). Aim of the study is to identify spatial correlations between different physical properties, especially thermal conductivity and density, and to use this correlation for large scale volumetric estimations. Modelling variograms for mineral resource estimation is a common practice but difficult for geothermal reservoirs due to an extreme sparseness of data. To circumvent this problem the variogram is constructed on dense data sets (e.g. core and/or outcrop) and transferred into the reservoir using density values calculated based on seismic sections. The validity of this approach depends on the correlation between thermal conductivity and density. If such a correlation is given a similar spatial correlation can be presumed. Finally the volumetric calculation of the petrophysical properties can be performed by Co-Kriging or Kriging with external drift. As an outlook the applicability of stochastic and inverse methods will be discussed |