Advanced workflow for multi-well tracer test analysis in a geothermal reservoir

Abstract

Interpretation of tracer tests in geothermal reservoirs is carried out by fitting the measured data either with simplified two-dimensional (2-D) analytical solutions or with complex numerical models. Available analytical solutions commonly only describe isotropic conditions in 1-D or 2-D, which is generally unsatisfactory to construct realistic reservoir models. Moreover, due to the large spatial and temporal scale of a tracer test in deep reservoirs, the concentration levels measured in the field are relatively low due to dispersion that complicates the assessment of breakthrough curve tailing and residence time. Fitting tracer data with fully resolving 3-D numerical models thus may be more appropriate, but even with a rich set of field data, reliable calibration is often compromised by the high computational effort and data hunger of such models. In this study, an advanced workflow is presented for evaluating tracer test data. Firstly, a 3-D analytical model in which solute transport is considered in anisotropic porous media is developed. Green's function method is applied to obtain a moving line source solution of the convection-dispersion-diffusion equation for solute transport in a 3-D porous medium. In addition, Green's function equation is analytically convoluted with a rectangular pulse function, which represents tracer injection. Secondly, the analytical model results are fitted to the tracer test data by Monte-Carlo simulations to obtain feasible ranges of flow velocities, as well as longitudinal and transversal dispersivities. Finally, the derived parameter values are implemented in a 3-D numerical model to evaluate the solute residence time in a large-scale reservoir. The results applied to field data from the Kizildere field in Turkey demonstrate that the workflow provides robust estimates of effective parameters from well-to-well data in complex reservoir systems with anisotropic flow paths. Thus, despite the higher effort of applying convolution and stochastic parameter estimation, the preceding analytical step of the workflow substantially eases numerical model set-up.