Geothermal Resource Assessment: An Experimental Design and Response Surface Method (ED and RSM) Approach

Abstract

Geothermal resource assessment is a highly uncertain exercise. The physics and thermodynamics of flow models are correlated with inherently uncertain geoscientific information to estimate the electric generation capacity of a geothermal resource. This has typically been done through volumetric methods calculating either the stored-heat or the mass in-place. Taking advantage of the ever-increasing power of computers, numerical geothermal reservoir simulations are inevitably going to be the standard for assessing the feasibility of generating electricity from geothermal resources. New Zealand is an example of this shift where geothermal developers are relying more on simulation results in decision-making. These advancements in geothermal reservoir simulation do not eliminate the uncertainty of the result and instead highlight the need for a more robust approach to probabilistic assessments when using the simulations. The experimental design and response surface method (ED and RSM) workflow provides such an approach. In ED and RSM, the uncertain parameters that affect the estimated generation capacity are tested through the reservoir simulation model. Multiple versions of the model are built using designed experiments that outline the combination and settings of the parameters tested. The various models or experiments are run and the generation capacity calculated from each run is used to fit a response surface model (proxy model). The response surface model, in the form of a polynomial, approximates the reservoir simulation model in terms of the parameters tested and the response estimated (generation capacity). Monte Carlo simulation is then applied to the response surface polynomial for a probabilistic assessment of the resource‟s potential for electricity generation. We have demonstrated the ED and RSM application on a synthetic geothermal dualporosity model and on the more complex Ngatamariki dual-porosity model. Results from these applications show that reservoir simulations can be used to generate probabilistic resource assessments. We also found that designed parameter combinations are sufficient to generate a response surface model that can be used effectively in place of the numerical model in Monte Carlo simulations. In addition, the results provide a more efficient approach to uncertainty management especially during the pre-development stage.