The Role of Uncertainty for Lumped Parameter Modeling and Opimization of Low Temperature Geothermal Resources
Author: Sven Klaus Scholtysik
Year: 2015
Supervisor: Ágúst Valfells, Hlynur Stefánsson & Halldór Pálsson
The project was conducted in cooperation with ÍSOR.
Abstract
With a growing population and ongoing industrialization, energy demand is rising on a global scale. Satisfying this demand in a sustainable way, while minimizing mankind's impact on climate change, is a significant challenge for our and future generations. One of the options that is apt to make a difference, is the use of low emission energy sources for heating purposes. One of these sources is Earth's ubiquitous geothermal potential. Understanding this potential and its limitations is of utmost importance.
Despite having a large impact on optimization and management of geothermal resources, the influence of uncertainty has not been studied extensively. Consequently, this thesis discusses the role of uncertainty in detail.
Using net present value maximization as the objective function, the sources of uncertainty are identified by creating a customized net present value model, which splits costs and benefits into different variables. These variables are analyzed for their tendency towards uncertainty. One of the influencing variables is the reservoir's physical reaction to production, as it allows forecasting of realistic exploitation values.
In order to test how much information is needed to produce good forecasting results, an initial lumped parameter model fit is obtained to identify the complexity of the best fit for four low temperature reservoirs in Iceland. In order to simulate decreasing uncertainty, the operational data is cut into smaller portions. By gradually extending the data range and iterative fitting, a development of the coefficient of determination is analyzed, finding that after ten seasonal cycles of data input, the model fit reaches a significant level of certainty. This time horizon can act as a stabilizing factor in the economic optimization and improve the accuracy of economic forecasting. Furthermore, the best fits do not show differences in model complexity for different levels of uncertainty.