Discrete Fracture Modeling of Hydraulic Stimulation in Enhanced Geothermal Systems (2010-2012)
Investigator: Mark McClure
In Enhanced Geothermal Systems (EGS), hydraulic stimulation is carried out by injecting water at high pressure into low permeability, typically crystalline rock. In most cases, the fluid injection causes slip on the preexisting fractures, enhancing their permeability and increasing well productivity. The simplest EGS arrangement is a two well doublet in which cool water is injected by an injector well, it heats up as it moves through the rock, and then is produced by a second well.
The potential size of the EGS resource in America and globally is very large. However it remains to be seen if EGS can be applied economically at a large scale. One challenge is that very high flow rates are needed per well to generate enough revenue to offset cost. Another is that as water flows through fractures from injector to producer, it may cool down the rock surrounding the fractures, resulting in premature breakthrough of cold water. A third is that the process of fracturing causes microseismicity. These small earthquakes are quite small, but have been as large as magnitude 3-4. These earthquakes are too weak to do serious damage, but they frighten and annoy local residents. At least one project, in Basel, Switzerland, was canceled due to a surprisingly large microseismic event.
The common link of the three problems listed above is that they relate to the process of hydraulic stimulation. For EGS to be successful, the process of stimulation needs to be mastered. Unfortunately, field experiments are costly and data is difficult to acquire. Computational modeling offers a way to run virtual experiments at very low cost. Modeling can be used identify promising techniques that could be tested in field demonstrations.
EGS stimulation modeling remains a relatively undeveloped field. There are major challenges associated with modeling EGS stimulation. It involves interacting hydraulic, thermal, mechanical, and thermoelastic processes. It is complicated by the need to account for the complex geometry of the preexisting fracture network. A handful of individual fracture pathways may take most of the flow between two wells.
This study uses discrete fracture networks to tackle the challenges of EGS modeling. Rather than regularly spaced, rectangular grid blocks, discrete fracture models are based on stochastically generated fracture networks which are discretized directly as an unstructured grid. Reservoir simulation is performed for the fluid flow in the fractures. In between time steps a boundary element method is used to calculate the location and magnitudes of the slipping fractures as well as how they perturb the stress on surrounding fractures. The rock permeability around the fractures is typically very low and is assumed zero in the model. On the short time scales of hydraulic stimulation, it is reasonable to use a one-dimensional heat conduction approximation for heat conduction towards the fractures. Therefore it is possible to avoid discretizing the rock around the fractures, dramatically improving the computational efficiency of the problem.To calculate thermoelastic stresses while using the 1D heat conduction approximation, a fast, accurate method was developed. Finally, we correlate the slip calculated on the fracture elements to microseismicity.
An example of a typical output from the model is shown in Figure 1. The thin horizontal line at the center represents a constant pressure injector. The thick lines are fractures that were stimulated. Even though all fractures have the same hydraulic parameters, the stimulated fracture network is very heterogenous and full of bottlenecks in flow. The uneven stimulation is caused by the stress perturbations of the slipping fractures.
Figure 1: Permeability of fractures following hydraulic stimulation. Thick lines are fractures that were stimulated, thin lines are fractures that were not.
In McClure and Horne (Stanford Geothermal Conference, February, 2010), the basis of the modeling work was described. It was shown that stresses induced by slip can dramatically affect the result of stimulation on a complex fracture network.
A paper was submitted for the Geothermal Resources Council Conference in October, 2010, that investigates strategies to reduce the magnitude and number of larger microseismic events during EGS stimulation. Modeling was focused on injection into a single, isolated fracture. An analytical expression was derived for the rate of advance of the stimulation that matched the numerical result. Various injection strategies were tested to identify how microseismicity might be reduced. It was found that reducing injection pressure over time reduced the magnitude and number of larger events. Putting a well immediately on production after injection reduced the post-injection events.
Many refinements to the current work are anticipated. Developing realistic realizations of the preexisting fracture network is a priority. It is planned to model the long term behavior of the reservoir, but that will require discretizing the rock between the fractures for calculation of thermal and poroelastic stresses and heat condution. Extension to three dimensions is an eventual goal. Ultimately, the goal is construct the most realistic simulator possible in order to identify techniques to improve EGS stimulation.