4.

1 Calcite scaling in well casings 

One of the common problems seen in geothermal wells is the deposition of calcium
carbonate or calcite, CaCO​3​, in the well casing, starting just above the flash horizon. It
is not uncommon for high-temperature geofluids to be close to saturation with respect
to calcite as they flow through the formation. The solubility of calcite, however, varies
inversely with temperature, so that it cannot precipitate from the geofluid merely
because of a decrease in temperature, other factors being constant. The other properties
of the geofluid that influence the solubility are:

● Partial pressure of carbon dioxide CO​2

● pH
 ● Salinity
● Calcium ion concentration.

The first two factors are interrelated; when the geofluid flashes in the well, the steam
that is released carries with it most of the CO​2​. This in turn causes the liquid pH to rise
dramatically, and results in the geofluid becoming supersaturated with respect to
calcite. Precipitation occurs immediately and can lead to severe narrowing of the
wellbore for several meters just above the flash horizon.
The chemical equilibrium reactions that control the process are [21]:

(4.48)
CO​2 ​(aq) ​⇔ ​CO​2 ​(vap)
+ ​HO
​ ​+ ​H+​ ​⇔ ​CO
HCO — (4.49)

3 2 2

Ca​++ ​+ ​2HCO —
​ ​⇔ CaCO ​↓ + ​HO ​+ ​CO ​↑ ​(4.50)

The loss of carbon dioxide (CO​2​↑​) upon flashing can be seen to cause the precipitation
of calcite (CaCO​3 ​↓​) from eq. (4.50). The fraction of CO​2​, relative to the original
amount dissolved in the brine, that leaves the brine with the flashed steam can be
found from [22]:

(4.51)
X​ = 1/​A
CO2​ [(1/​x​) ​— ​1] ​+ ​(1/​A​)

where ​A ​is a separation coefficient [23] and ​x ​is the dryness fraction of the two- phase,
steam-liquid mixture. The separation coefficient is a function of temperature and the
salinity of the brine, as shown in Figs. 4.11 and 4.12. The dryness fraction
 ​ ​ (1/A) as a function of temperature for various molalities.
Fig. 4.11 Log10
 ​ ​ (1/A) as a function of molality for various temperatures.
Fig. 4.12 Log10

can be calculated from eq. (4.31). For example, if a 1 M brine at 200​° ​C undergoes a
12% flash, then 98.5% of the CO​2​ will end up in the vapor phase.

4.2 Reservoir modeling and simulation 

A reservoir simulator is a computer code that embodies all of the essential properties of
the formation, the geofluid, and the environment together with the physical and
chemical laws controlling their interactions to allow the extrapolation of the
performance of the reservoir from the present time to a future time, usually several
decades away.
The information derived from such a code is valuable to those interested in
developing and financing a power plant. Often a reservoir model is the only predictive
tool available at the early stages of development to assure potential investors that their
money will be wisely spent. It is also a requirement for most feasibility studies when
considering a new project or the expansion of an existing one.
Nowadays, with the widespread use of personal computers with powerful
capabilities once reserved for mainframe or mini-computers, it is far easier to conduct
reservoir modeling studies. However, the relative ease with which such studies can be
performed does not mean that the results can be viewed as any more reliable.
Reservoir simulations require a solid understanding of the geothermal system, based
on the geoscientific exploration results including the results of the first few deep wells.
Unless this information has been crafted into a model truly representative of the
geothermal system, one that may extend for tens of square kilometers on the surface
and several thousand meters below the surface, and unless suitable methods of solution
for complex equations have been devised, then the results of the simulation will have
little predictive value.
Reservoir modeling and simulation is the subject of annual symposia, and numerous
technical articles. The reader is referred to the following sources for
detailed presentations: Refs. [1, 2, 24–38]. Several universities, companies and
governmental laboratories have developed their own versions of reservoir simulators,
usually identified by unique acronyms. The private ones are usually proprietary and all
computer programs periodically evolve to newer versions. In Sect. 4.5.6, we list some
of the more commonly used geothermal reservoir simulators, briefly highlight their
features, and include references for details on each one.

4.2.1 Input

A reservoir model must be built on a proper conceptual model of the field. If the
conceptual model has flaws, as often happens in the early stages of field development,
the reservoir model will reflect these errors. For this reason, it is recognized that the
reservoir model must be updated as the conceptual model improves.
The input includes as much of the following information as possible:

● Geology: rock types and ages, outcrops, surface topology, faults, calderas,
craters, eruptive centers, discontinuities
● Hydrology: rainfall patterns, surface water flows
● Geophysics: surveys of electrical resistivity, gravity, magnetics, temperature
gradients, heat flux
● Geochemistry: hot and cold spring compositions, flow rates and temperatures,
estimated reservoir temperatures
● Well data: locations, depths, casing design, pressure and temperature profiles
both static and flowing, productivity curves.

All of this data permits the construction of the conceptual model which should
indicate the areal extent of the reservoir, the areas of natural fluid discharge and
recharge, the natural heat flow from the reservoir by conduction through rock and by
convection through thermal manifestations, and an estimate of the vertical extent of the
geothermal system.

4.2.2 Architecture
Since it is virtually impossible to write down analytical equations to describe all the
physical and chemical processes going on in a geothermal reservoir and to solve them
analytically in a closed form, all reservoir models rely on some type of finite
difference or finite element numerical technique. The field is divided into discrete
cells, both horizontally and vertically, and the governing equations of mass and energy
conservation are written in finite difference form for each element of the matrix. The
conditions on the boundaries of each cell provide the connections among the cells and
permit the solution to extend over the entire reservoir.
The choices of the cell boundaries are arbitrary and may be selected for
computational convenience, but usually they coincide with major structural features
such as faults or permeability barriers. The size of the cells is also arbitrary and the
grid usually is made finer in the neighborhood of well-defined features, such as wells,
where data is plentiful and reliable. Outlying areas can be represented
by large blocks having average properties without greatly influencing the results of the
model. The geometric shape of the cells is also arbitrary, ranging from simple squares
or rectangles to circles or irregular polygons. It is often convenient to situate each well
within its own element.
The vertical extent of the reservoir is modeled by layering the matrix into several
distinct horizons. For young fields with little production history, it is not uncommon to
have six or more layers covering, for example, the surface and the atmosphere, the
impermeable cap rock, a liquid-dominated zone, a particular rock formation, a deep
permeable zone, and the basement rock. In more developed fields with a long
production history, models are much finer in structure and can have dozens of layers
[39].

4.2.3 Calibration and validation

Once the governing spatial- and time-dependent physical and chemical equations have
been written, and the code created along with the method of solution, it is then
necessary to calibrate and validate the model. This crucial step is needed to guarantee
that the model (and its solution technique) is reliable, stable, and correctly represents
the state of the reservoir going back in time perhaps as much as 250,000 years – from
the unexploited, natural state of the reservoir to the present. In this way, the model can
be relied upon to accurately model the ancient natural state of the system, as well as
the present, unexploited state of the system, and to offer the modeler confidence in
making predictions about the future state of the system.
This is a very tedious computational process. Imagine a mathematical model
involving strongly-coupled, non-linear, partial differential equations represented by a
computer code with hundreds of input parameters, many of them ill-defined owing to
insufficient data. The set of equations is said to be mathematically under- constrained,
having many more parameters than can be determined or inferred from field
measurements. A solution in such a case will not be unique, there being many possible
conditions that will satisfy this set of equations. Judgment is needed by the user of the
code to make good choices for the adjustable parameters to arrive at reasonable results.
In what is called the “manual” method, the parameters first are adjusted by the
reservoir engineer to give the best fit possible to the current state of the system. The
two parameters most often used are the permeabilities of the various rock formations
and the boundary conditions [40]. Then the code is run from the natural state forward
in time and the physical characteristics of reservoir are computed at large time steps.
The evolution of the field is thus depicted. Stability of the code requires that eventually
the system should reach a steady state that conforms to the observed present state. If
the results of this computation are unsatisfactory for any reason, then the modeler must
adjust the parameters and re-run the simulation until satisfactory results are obtained.
This process can be partly automated in a sense by having the reservoir properties
adjusted by a programmed subroutine within the code to minimize the difference
between the code output and selected observed data. The automatic method is called
“inverse” modeling, but still requires the intervention of the modeler in choosing
the properties that will serve as criteria of the goodness of the fit and the magnitude of
the errors that are acceptable. Inverse modeling is more demanding on computer speed
and memory than manual modeling, and may require parallel processing using several
personal computers.

4.2.4 History matching

The first use of the model is to match the data obtained from the early wells. The model
should be able to reproduce the observed changes in the reservoir over the period of
production and injection. When this is done during the feasibility stage of a new field,
such data may be scanty, but the model still must be able to match it. This usually
means another adjustment of the parameters in the model, and a recalculation of the
natural state model, followed by more iterations to bring the model into agreement
with all the data. As more production and injection data become available, the model
should be revised accordingly.
Here again, this process of history matching can be done either manually or
automatically. This cycle of periodically updating the reservoir simulator is tedious,
time-consuming and expensive, but absolutely essential if one is to have any
confidence in projecting the behavior of the field over many years of operation.

4.2.5 Use of the model

A validated reservoir model may be used to forecast the behavior of the system when it
is subjected to various exploitation scenarios. This involves running the model for
future times after having placed projected future production and injection wells at
specific locations in the field, ie, mathematically inserting fluid sinks (production
wells) and sources (injection wells) into the reservoir. The desired outputs from the
model include the evolution of pressure and temperature, the formation of vapor zones,
the movement of chemical and thermal fronts through the reservoir, the onset of
chemical and thermal breakthroughs from injection wells to production wells, the level
of electrical generation that can be supported for a specified number of years, typically
20–30 years, and the number of replacement wells that must be drilled over the years
to maintain the electrical output.
Given the complexity of the rock-fluid interaction in an imprecisely known
geometric structure, it is easy to see how such a modeling process is subject to
considerable uncertainty. Furthermore, since geothermal regions are dynamic and
subject to geologic forces that can change the structure of the formation, opening new
fissures or closing old ones, moving fault blocks laterally as well as vertically, the
assumed constant geometric structure may be altered by natural forces. This could play
a role when the simulator is used over the millennia of the natural state computation,
during which the reservoir is assumed to have a fixed configuration. Indeed, the field
may be altered by geologic forces during the anticipated life of the power plant, as has
been the case at some geothermal power plants around the world. Notwithstanding
these difficulties, reservoir models, after they have been validated and calibrated
against whatever data is available and periodically updated, have proved to be
 reasonably reliable tools for simulating the state of geothermal systems. An excellent
review of over 125 field simulations based on questionnaires
filled out by companies and individuals can be found in Ref. [40].