15.01.2015 Views

Microseismic Monitoring and Geomechanical Modelling of CO2 - bris

Microseismic Monitoring and Geomechanical Modelling of CO2 - bris

Microseismic Monitoring and Geomechanical Modelling of CO2 - bris

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

calibrate. This model has been calibrated with over 200 different core measurements, <strong>and</strong> has been<br />

found to be remarkably consistent.<br />

I have applied this model to simple geomechanical models in order to determine the sensitivity<br />

<strong>of</strong> seismic properties to stress path effects. The results have demonstrated the potential to diagnose<br />

what stress path a reservoir is following using 4-D seismic techniques. By diagnosing the stress path,<br />

geomechanical models can be calibrated, allowing the risk <strong>of</strong> fracturing to be determined. I have also<br />

demonstrated how anisotropy could potentially be used as an indicator <strong>of</strong> reservoir compartmentalisation.<br />

Having developed <strong>and</strong> demonstrated the workflow to move from geomechanical modelling to seismic<br />

predictions, I apply the concept to the real example <strong>of</strong> the Weyburn field. I have generated a simple<br />

model that represents the major features <strong>of</strong> the reservoir. The initial material parameters used to<br />

populate the model were derived from core sample measurements. However, the predictions from this<br />

model do not provide a good match with either microseismicity or anisotropy observations. It is a<br />

well known (<strong>and</strong> yet <strong>of</strong>ten ignored) fact that mechanical tests on core samples do not, for obvious<br />

reasons, include the effects <strong>of</strong> large scale fractures on the overall properties <strong>of</strong> a material. Upscaling<br />

commonly finds that the rock mass is s<strong>of</strong>ter than determined by core tests, <strong>and</strong> this is particularly true<br />

for heavily fractured rocks, such as the Weyburn reservoir. When the Weyburn model is recomputed<br />

with a smaller Young’s modulus for the reservoir, a much closer match is found between model<br />

predictions <strong>and</strong> observations.<br />

There are many free parameters that can be varied in a geomechanical model, all <strong>of</strong> which can<br />

influence the result. We generally do our best to constrain the information put into such a model<br />

using many sources to aid model population, from 3-D controlled source seismics, borehole logs <strong>and</strong><br />

core work. However, this information does not directly correspond to the information that is needed<br />

- i.e. the poroelastic <strong>and</strong> plastic response <strong>of</strong> the rock mass, as a whole (not a limited, core-sized<br />

sample), to the relatively large <strong>and</strong> long-term stresses applied during CO 2 injection (as opposed to<br />

the low-magnitude, short duration stress applied by an acoustic or seismic wave). These measurements<br />

provide proxy information for what we really need. As such, each parameter in the model has a degree<br />

<strong>of</strong> uncertainty to it.<br />

The question then must be: how should we deal with this uncertainty in order to have any kind<br />

<strong>of</strong> confidence in model predictions One option is to use a stochastic method, where a probability<br />

function is assigned to each parameter, <strong>and</strong> the resulting probabilities for the results are computed.<br />

However, even such a method must first assign probabilities to the input parameters, which does<br />

not get away from the original problem, in that the probability function will be based on proxy<br />

measurements <strong>and</strong> will not directly represent the required parameter. Therefore we must have a<br />

method for determining which models provide the most appropriate results, a decision which must<br />

be based on comparison with observations in the field. A number <strong>of</strong> observations might be made<br />

with which to groundtruth geomechanical models, such as surface deformation or borehole tiltmeters.<br />

In this thesis I have used observations <strong>of</strong> induced seismicity <strong>and</strong> anisotropy to constrain my models,<br />

<strong>and</strong> have demonstrated that some <strong>of</strong> the information used to construct the original model, in this<br />

case the Young’s modulus provided by core sample measurements, does not actually do a good job <strong>of</strong><br />

171

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!