Abstract:
Geothermal wells are the lifeblood of a geothermal development, warranting extensive
testing and monitoring. The general term well testing is used in relation to a wide array
of testing programmes and techniques, including pressure transient analysis (PTA),
a technique which is under-utilised by the geothermal industry due to the fact that conventional
PTA analytical models often cannot match eld datasets from real geothermal
wells, or give results that do not make sense. This is primarily due to the high temperatures
and two-phase conditions of geothermal systems. These can only be adequately
represented by numerical modelling, however there has been a lack of numerical tools or
software for geothermal PTA.
This research divides naturally into two parts: practical issues related to collection of
geothermal PTA eld datasets, and numerical techniques for analysis of those datasets.
The majority of the practical issues can be avoided or minimised with careful test design
and adequate data logging, and a set of recommendations forms part of this thesis.
A core component of this research is the numerical PTA framework, a range of
guidelines for setting up the numerical simulations. It has been implemented in Python,
using PyTOUGH to control the numerical simulator TOUGH2, though it is applicable
to other simulators. It has been demonstrated on a wide variety of geothermal eld PTA
datasets, proving superior to analytical models. It can account for a variety of e ects,
including non-isothermal conditions during injection, and can model well behaviours
beyond the capability of conventional analytical models, including sudden changes in
permeability during the pressure transient.
The numerical PTA framework has been applied to a variety of eld datasets from
geothermal wells, including assessment of stimulation due to de agration and thermal
stimulation during injection. A range of reservoir-scale structural features have been
identi ed, including a single linear impermeable boundary, channel boundaries of various
widths, and reservoirs with di erent fractional dimensions. The reservoir-scale structural
features can be highly signi cant to the conceptual model of a eld, and this is also
explored.