| Abstract |
Models for simulating heat and mass transfer in sedimentary basins require petrophysical input data (e.g. permeability and thermal conductivity). Knowledge of the spatial distribution of these system parameters is typically very limited. Measurements come from different sources and are distributed on multiple scales, being derived from e.g. well-logs, outcrop analogues, plugs, or thin sections. Remarkable databases of petrophysical properties have been maintained in the past. Projecting these measurements onto reservoir models requires, for example, geostatistical property modeling, upscaling and possibly also downscaling. There are many well-established property modeling and upscaling techniques for petrophysical properties in hydrocarbon reservoirs. However, it is less well understood how these properties should be modeled and upscaled in geothermal reservoirs. Different physical properties can be associated with different representative volumes (REVs), depending if they are needed to model heat transport or mass transport. Yet, geothermal reservoirs have a much tighter margin to be economically profitable, so being able to model and upscale petrophysical properties in such reservoirs is of particular importance to forecast heat in place and energy extraction reliably.As a starting point, we address this challenge by discussing and reviewing different property modeling and upscaling techniques for thermal conductivity. These include techniques based on classical volume averaging and homogenization as well as statistical scaling methodologies which can lead to propagation of uncertainty across scales. Although it is possible to utilize general principles from the upscaling of classical hydraulic properties such as permeability, there are some important differences. A simple and frequently used approach for assigning petrophysical properties to a reservoir model is assigning an average value to units of the same geological age. However, when modeling the distribution of thermal conductivity, a more robust approach should be geologically-based and include information of the facies, diagenesis and local variations in mineral content. This approach has to be generalized by modeling the spatial distribution of thermal conductivity using geostatistical and inverse techniques, eventually based on a robust dynamic data-assimilation framework, grounded on a-priori geological information and the way key (statistical) moments are transferred across scales. |