PROCEEDINGS, TOUGH Symposium 2012 Lawrence Berkeley National Laboratory, Berkeley, California, September 17-19, 2012 - 1 - BASIN-SCALE GEOTHERMAL MODEL CALIBRATION USING ITOUGH2 Lynn B. Reid 1,2,3 , J. Florian Wellmann 2 1 NTEC Environmental Technology, Subiaco WA 6008 Australia 2 CSIRO Earth Science and Resource Engineering, Kensington WA 6151 Australia 3 School of Environmental System Engineering, Uni. Western Australia, Crawley WA 6009 Australia lynn.reid@uwa.edu.au ABSTRACT We showcase how the application of parameter- estimation and sensitivity-study methods implemented in iTOUGH2 improved model calibration to measured data and provided insight into model uncertainties and data outli- ers. The methodology was applied to a basin- scale conductive heat model of temperature in the Perth Basin, Western Australia. MOTIVATION AND INTRODUCTION Geothermal resource estimations require a trade- off between conceptualizing the complicated geological system with a finite numerical repre- sentation, and estimating these model parame- ters from scarce data. There is often a high degree of uncertainty about the heat flow param- eters used in the simulation, especially for large- scale systems with poorly detailed boundary conditions (see, e.g., Kohl, 2003). A suitable input parameter set may be estimated by first performing a forward simulation of temperatures, and second by comparing simu- lated to measured values in wells. A common approach is to simply compare values by “eyeballing”, for example by plotting simulated next to measured values along a temperature log (Saibi, 2011), or by comparing a sum of simu- lated mismatch across all data points (Reid et al., 2012a). The input parameters are then adjusted by trial-and-error until a reasonable fit is obtained. Although this method seems to be very “ad-hoc,” it is usually possible to obtain acceptable results with a reasonable amount of manual calibration steps. For purely conductive geothermal models, this approach may be adequate, as the forward simulation is approxi- mately linear in its response to the parameters and the sensitivities of the single model param- eters can be reasonably well determined. This simple manual calibration method can be automated with several more sophisticated model calibration methods. The advantages of automatic calibration are a requirement to math- ematically define what constitutes a “good” fit, systematic analysis of the sensitivity of the computational model to both the parameters and the calibration data, an indication of the adequacy of the conceptual model, and calibrated parameters which hopefully provide an adequate fit to measurements—of course all at the expense of computational complexity. We do not claim that automatic calibration methods produce a “true” or unique model (Moore and Doherty, 2005), but the approach, when applied carefully, will produce a well-considered model. In this work we apply the iTOUGH2 program (Finsterle, 1999) to automatically calibrate a complex geothermal model using the PEST protocol (Finsterle and Zhang, 2011). The conductive forward simulations are performed in SHEMAT (Clauser and Bartels, 2003), as multi- phase flow is not considered at this stage. We utilize the stand-alone capabilities of iTOUGH2 to estimate thermal rock parameters and bound- ary conditions. Our aim is to produce a simpli- fied basin-scale model to act as a basis for future detailed reservoir-scale simulations representing more complex physical heat transport mecha- nisms, such as convection and groundwater advection. GEOTHERMAL MODELING PROJECT We developed three-dimensional models of the geology and the conductive temperature regime of the entire Perth Basin in Western Australia, with an extent of nearly 800 km from north to south and 150 km east to west. The enormous physical scale required many simplifications to integrate the three-dimensional geological model and discrete geothermal simulation.