Geothermics 38 (2009) 6471

Contents lists available at ScienceDirect

Geothermics
journal homepage: www.elsevier.com/locate/geothermics

Pilot plant experiments at Wairakei Power Station

Kevin L. Brown a, , Lew G. Bacon b
a
b

GEOKEM, P.O. Box 92019, Swanson, Waitakere City, New Zealand

Eastland Environmental Services, Opotiki, New Zealand

a r t i c l e

i n f o

Article history:
Received 5 June 2008
Accepted 25 November 2008
Available online 15 January 2009
Keywords:
Geothermal
Silica scaling
Silica sols
Condensate corrosion
Steamline scrubbing
Arsenic removal
Wairakei
New Zealand

a b s t r a c t
In the mid-1990s, several pilot plants were constructed at Wairakei to either improve the operational and
economic performance of the power station or to mitigate the environmental effects of discharges to the
Waikato River. The results of the following investigations are discussed: (1) uid ow dynamic effects
on silica scaling; (2) production of silica sols of predetermined particle size to evaluate the potential for
generating commercial grade silica products; (3) use of sulfur oxidising bacteria for the abatement of
dissolved hydrogen sulphide in cooling water; (4) removal of arsenic from separated geothermal water; (5)
steam line condensate corrosion; and (6) measurement and modelling of steam scrubbing in Wairakeis
long steamlines.
2008 Elsevier Ltd. All rights reserved.

1. Introduction

2. Silica scaling under controlled hydrodynamic conditions

Wairakei was the rst liquid-dominated geothermal eld that

was developed for power production, and in its early days had to
pioneer a number of techniques to solve problems that arose during
the normal running of the station and boreeld. In the mid-1990s,
a number of pilot plants were constructed to investigate ways of
improving the operational and economic performance of the power
station or mitigating the environmental effects of discharges to
the Waikato River. As such, part of the work undertaken was not
directed at science but at outcomes. Some of the pilot plants have
an on-going intellectual property component and are only briey
discussed. A plant was also established to investigate if a high-value,
commercially acceptable, silica product could be produced which
could offset the cost of extraction and provide a commercial return.
This paper includes material taken from Dunstall and Brown (1998),
Brown and Bacon (2000), Brown and Dunstall (2000), and Dunstall
et al. (2000).

2.1. Silica scaling

Corresponding author. Tel.: +64 9 833 3331; fax: +64 9 833 3331.
E-mail address: kevin@geokem.co.nz (K.L. Brown).
0375-6505/$ see front matter 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.geothermics.2008.11.004

In uids supersaturated with respect to amorphous silica the

mechanism responsible for the majority of silica deposition is the
preliminary formation of colloids, and their subsequent precipitation on to equipment surfaces as a voluminous, sometimes hard,
but often porous scale. These deposits reduce heat transfer in heat
exchangers, increase pressure drop in pipes and, in severe cases, can
cause complete blockages in sections of the system (e.g. Gunnarsson
and Arnrsson, 2005). Although silica scaling is a widely observed
problem in geothermal operations, the actual mechanism by which
the silica colloids are transported from the bulk uid to the scaled
surface is poorly understood at a molecular or colloidal level.
Experiments were set up with the objective of obtaining a more
fundamental understanding of the silica scaling mechanism as it
applies to geothermal equipment.
Previous work (e.g. Weres et al., 1982) had primarily emphasized the polymerisation kinetics of silica and how they are affected
by pH, other ions in solution, temperature, and degree of supersaturation. Consequently, although there was not sufcient data
to condently predict an exact polymerisation curve, the general
principles were known by the mid-1990s. However, relatively little
research had been undertaken to investigate hydrodynamic effects
on silica scaling.
It had been empirically observed that uid ow structure can
inuence silica scaling with unusual deposition being observed

K.L. Brown, L.G. Bacon / Geothermics 38 (2009) 6471

65

Fig. 1. Schematic diagram of the test section in the silica scaling pilot plant at Wairakei. 500:500-mm diameter pipe.

near bends, valves and other items which disturb or disrupt the
ow. Some years previous to the experiments described here, a
simple study to determine the hydrodynamic inuence on silica scaling was conducted in the wastewater drains at Wairakei
(Garibaldi, 1980). Two cylinders and a horizontal at plate were
exposed to the ow for several weeks, providing qualitative evidence that uid hydrodynamics played an important role in the
deposition process. For example, the stagnation point at the front
of an exposed cylinder showed a high rate of deposition, and less
silica was observed at points 90 to the ow axis, where the velocity is highest. The morphology of the silica scale was also found
to be dependent on the uid ow, with cellular silica structures
in areas of recirculation and more needle-like structures in welldirected ow. Distinct zones of deposition were also seen on the
at plate, and these appeared to relate to laminar, transitional and
turbulent ow regions. However, the ow conditions within the
drains were difcult to characterise and this placed a limitation on
the interpretation of the results.
In another study (Pott et al., 1996) the deposition of silica onto
a at plate in laminar ow was modelled numerically. Attractive forces acting between particles and the plate were modelled
simultaneously with uid ow, using a nite difference method.
This study indicated that the deposition rate was dependent
on particle size, with small particles depositing more quickly,
and that the rate would initially be greatest at the front of the
plate. It was predicted that an initial build-up of silica scale at
the front of the plate would change the shape of the laminar
boundary layer. This could lead to ow separation and an earlier
transition to turbulent ow, hence increasing colloidal deposition.
2.2. Pilot plant design
The pilot plant therefore attempted to identify the effects of ow
characteristics on scaling by controlling both silica colloidal particle size and uid ow conditions simultaneously in order to gain a
fundamental understanding of the forces involved in an individual
colloid particle moving from the bulk solution phase and becoming attached to the wall. The hydrodynamically well-examined and
prescribed geometry of a at plate in a parallel laminar ow was
the rst ow situation to be investigated, followed by a vertical
cylinder.
To provide a well-characterised hydrodynamic ow, a water
tunnel was constructed which was large enough that the test samples placed within it would not seriously impede the ow. This
restriction necessitated a tunnel diameter of 200 mm. The range

of uid velocities required was from 0.5 to 3 m/s, which required

a geothermal ow rate of 56336 tonne1 /h. Because such a large
quantity of geothermal uid was required to produce the ows in
the test section, a design was used where treated uid was recirculated using a pump. A small fraction of the recirculated uid
was drawn off from this circuit (210 t/h) and replaced by freshly
treated geothermal brine. The average residence time for treated
uid in the recirculating loop was 735 min, depending on the
make-up ow rate. A schematic of the test system is shown in
Fig. 1.
A uniform velocity is desired across the test section, to provide
constant conditions across the entire width of the test piece. This
was achieved by allowing the uid to settle in a 500-mm diameter pipe (at low velocity) and then accelerating the uid quickly
into the test section. Provision was made to insert and secure models and test plates in the test section through four large (50-mm
diameter) access ports at 90 spacing around the perimeter. Ten
smaller access ports (12-mm diameter) were included in two planes
along the axis of the test section to provide access for a pitotstatic probe. All of the access port plugs were smoothed to conform
to the inner wall of the test section to prevent disruption of the
ow.
Silica particle size control was obtained by manipulating
the nucleation temperature. The particle size distribution was
monodisperse, typically with a standard deviation of about 6 nm.
A plate heat exchanger provided initial cooling of uid from the
Wairakei reinjection line. This uid was then passed to a hold-up
tank to allow for silica particle growth to stabilise at the required
size and then the uid was passed into the recirculating section
of the plant. Colloid size was measured using photon correlation
spectroscopy (Pecora, 1985).
The whole pilot plant was fully instrumented for ow control, data collection, temperature, etc. Use of programmable logic
controller instrumentation allowed a constant uid velocity to be
maintained within the test section even if silica scale were to signicantly alter the pressure drop in the recirculating loop during
the experiment. Local dynamic pressures and temperatures within
the test section were also measured; pressure data were obtained
using a transducer connected to a 4-mm diameter pitot-static tube
which could be inserted through any of the access ports in the side
or top of the test section.

Tonne (t) = 1000 kg.

66

K.L. Brown, L.G. Bacon / Geothermics 38 (2009) 6471

Table 1
Results of plate experiments in the silica scaling pilot plant at Wairakei.
Test no.
a

2
3
4
5
6a
7a
8
9a
a

Silica size (nm)

Fluid velocity (m/s)

Experiment duration (days)

50
22
29
34
35
Not measured
12
95

1.2
1.2
1.2
2.0
2.0
2.0
2.0
1.0

14
31
17
15
<7
<7
43
8

Experiment halted early.

2.3. Initial test results

Initial tests conrmed that the velocity prole in the test section was sufciently uniform and that additional ow-straighteners
were unnecessary. The rst test specimen was a 100-mm long stainless steel plate, sharpened to a knife edge at the upstream end.
This plate was exposed to a ow velocity of 1.2 m/s and a silica
particle size of approximately 8.5 nm for a period of 20 days. Very
little scaling was observed on this plate, and certainly none of the
voluminous silica deposition normally associated with geothermal
applications. This test conrmed that the plate mounting method
was performing satisfactorily and that the control system could
provide constant ow conditions throughout the test. The absence
of visible scaling on this plate is consistent with previous experiments that had shown that very small particle size colloids display
very low silica deposition rates. This is in contrast to the theoretical
results deduced by Pott et al. (1996).
2.4. Further plate tests
After the rst test the plate design was modied. Subsequent
plates were 250 mm long, made of mild steel, and included, on one
side of the plate only, a trip wire 5 mm behind the sharpened leading edge. The purpose of this wire was to cause the boundary layer
to become turbulent, so that a comparison could be made with the
naturally developing laminar boundary on the other side. As well,
the particle size was increased to enhance the likelihood of scaling. A series of tests (Table 1) was conducted with the new plates,
covering a range of particle sizes and ow conditions. The at plate
tests were designed to cover the four options of smaller and larger
particle sizes together with lower and higher ow rates. These ow
rates could be carefully controlled, but the size of the particles was
less well regulated.
Results obtained with the new plates were encouraging, but also
somewhat surprising. Those of Test 2, the rst test with the trip wire
and the mild steel plate, were unexpected. Little scaling was seen
over the rst 40 mm of plate length with substantial scaling further downstream. Both sides of the plate had the same appearance
(i.e. no noticeable effect due to the trip wire), and the plate showed
considerable scaling whereas the inside of the pipe in the test section appeared clean. Test 3 was conducted using plate material
sourced from a different supplier, and gave quite different results,
with much less scaling in fact, no noticeable scaling at all. At this
time it became clear that Test 2 had showed unusual behaviour and
a query to the manufacturer revealed that the mild steel had a very
thin coating of sprayed zinc for protection. This coating had been
partly removed when the knife edge was machined and, in this area,
little scaling had occurred. Where the zinc remained intact, deposition was much greater. The untreated carbon steel walls of the pipe
test section, which was coated with normal mill scale, remained
relatively silica free. In order to conrm that the zinc was responsible for the greatly increased scaling rate the samples installed in all
subsequent tests had the zinc coating removed from half of the plate

Fig. 2. The six trial cylinders from the silica scaling pilot plant at Wairakei. From left
to right: large particles, low velocity; large particles, high velocity; medium particles,
low velocity; medium particles, high velocity; small particles, low velocity; small
particles, and high velocity.

on both sides. Test 4 was the rst test performed in this way, and
conrmed that the presence of zinc on the surface greatly increased
the rate of silica scaling.
Despite changes to the velocity and particle size in subsequent
tests there was no observable inuence of hydrodynamics in the
scaling process. Although it is difcult to quantify with the experiments having variable length of exposure, there appears to be a
correlation of scaling rate with particle size. Smaller particles cause
less scaling than larger ones.
2.5. Vertical cylinder tests
A series of vertical cylinder tests followed the at plate experiments. In these tests the mild steel cylinders were 25 mm in
diameter, and sized to t exactly across the whole diameter of the
test section. Six experiments were completed with these cylinders
which covered the variables of large (120 nm), medium (70 nm)
and small (15 nm) particle size, together with fast (2.5 m/s) and
slow (1 m/s) uid velocity. All of the cylinder experiments lasted
approximately the same time; i.e. about 3 weeks.
The results obtained in the cylinder experiments were different from those of the plate tests. The scales deposited were much
harder than those deposited on the at plates, and a hydrodynamic
inuence was immediately obvious. Most striking was that no visible scaling occurred along the whole length of the stagnation line
(Fig. 2). This clear area was about 3 mm wide. The silica free area
on the stagnation line was wider for lower velocities with the same
particle size and also slightly wider for smaller particles at the same
velocity. On either side there was an area of picket fence silica
scaling, with the fences being aligned parallel to the length of the
cylinder, or perpendicular to the ow stream lines, and pointing
into the ow. This type of scaling was evident from the edge of the
stagnation line, to a point about 90 around the cylinder from the
forward facing side. The back half of the cylinder, which was facing
away from the ow direction, was essentially free of scale.
There was also a very marked effect due to particle size: the
larger the size the greater the degree of scaling. Silica deposition
at the smaller particles sizes was almost non-existent (Fig. 2). The
scales deposited on the cylinders were much more like those commonly seen in geothermal elds than those observed on the at
plates.
We were not able to discern any difference in silica scaling
between either ends of the cylinders, which indicates that there

K.L. Brown, L.G. Bacon / Geothermics 38 (2009) 6471

67

is a negligible gravity effect in these experiments. Where the ow

was disturbed close to the test tunnel walls, the silica scale was
aligned in a pattern that indicated it was forming perpendicular to
the ow stream.

2.6. Conclusions
The experiments indicated that there was a denite colloid particle size effect, as well as an unmistakable hydrodynamic effect
in the cylinder tests. Silica scaling occurs at an individual colloid
level only when a colloid is transported from the bulk uid to a
scaling surface. This will require transport of the colloid through
the boundary layer adjacent to the surface. We considered a number of possible mechanisms for this transport process: electrostatic,
diffusion, inertial and/or gravitational forces.
We conclude that inertial forces were probably responsible for
the colloid transport through the boundary layer. Individual colloid particles will have a velocity distribution in the uid due to
Brownian motion. If the momentum is sufcient, the particles can
penetrate the boundary layer to attach to a surface. The effect will
be proportional to the mass and velocity of the particles. The mass
of a colloid particle is proportional to its radius cubed, and so there
should be a very large particle size effect, which we do observe. This
would also explain the smaller silica-free stagnation line at higher
velocities. The lack of scaling on the downstream side of the vertical cylinders can be explained by a reduction in the inertia of the
particles when the ow separates on the sides of the cylinder. In
this area, the uid velocity is greatly reduced.

3. Manufacture of silica sols from separated geothermal

water
3.1. Silica sols
The deposition of silica in boreeld and power station installations tends to limit the efcient use of geothermal energy. One
possible solution to this problem is to remove the silica from the
geothermal brine, which would then allow further utilisation of
heat not normally available. Our aim was to produce a high-value
commercially acceptable silica product that would not only offset
the cost of extraction, but also provide a commercial return. A further requirement was that any process did not constrain normal
power station operation.
Silica sols are concentrated solutions of colloidal silica in which
the colloid particles are constrained within narrow size limits (i.e.
monodisperse), and where impurity levels are appropriate for the
nal application. A further requirement is the necessity to produce
monodisperse sols with different average particle diameters. Typical uses of such sols are in refractory cements, polishes for silicon
wafers, photographic coatings and as binders for catalytic materials.
Initial bench top experiments were conducted to determine
how to measure particle size, control particle size distribution, stabilise the particles to halt agglomeration, concentrate the colloidal
solution, remove contaminants, and determine typical scale-up
parameters for a larger pilot plant. These parameters were then
used to design a small pilot plant that would provide enough sample
material for commercial evaluation and provide sufcient operational data.

Fig. 3. Schematic diagram of the silica extraction pilot plant at Wairakei.

ltration (UF) to concentrate the sol, incorporating dialtration2 to

remove contaminants. Initial eld experiments at Wairakei had elucidated the silica polymerisation kinetics, which allowed the design
of the colloid growth section of the plant. The use of UF is commonplace in New Zealand in the dairy milk processing industry and this
experience facilitated the selection of UF membranes.
The essential elements of the pilot plant consisted of:
a. a supply of separated geothermal water which was undersaturated with respect to amorphous silica, taken from the
injection pipeline;
b. a heat exchanger to reduce the temperature of the separated
geothermal water from approximately 125 C to the selected silica sol nucleation temperature;
c. a series of temperature controlled ageing tanks to enable control
of ageing temperatures during colloid growth;
d. a 100-m lter to remove suspended solids;
e. a three-stage continuous UF system to concentrate the preformed colloidal silica to a concentration of >30% silica by
weight;
f. dosing pumps to enable pH adjustment and the addition of stabilising agents;
g. a fresh water supply to enable the removal of geothermal salts;
and
h. ancillary equipment to enable mixing and circulation of chemicals for sterilising and cleaning the UF lters and pipe-work.
A schematic of the pilot plant is shown in Fig. 3. A comprehensive computer-based process control system enabled all process
variables to be selected, measured and recorded.
3.3. Silica particle size
As in the hydrodynamic experiments (see Section 2), the colloid
particle size was determined by the nucleation temperature. This
initial temperature drop to the nucleation temperature provided
a degree of silica oversaturation that causes nucleation of individual silicic acid molecules to eventually form polymerised colloids.
The chemical driving force for nucleation is the degree of supersaturation, and therefore the larger the degree of oversaturation
the greater the number of nuclei formed. Ageing and further gradual cooling by careful temperature control allows growth of the

3.2. Method and equipment

The conceptual design of this pilot plant was based on initial
temperature control to regulate the particle size followed by ultra-

2
A technique that uses dilution and additional ultraltration to remove, or lower,
the concentration of salts and contaminants from solutions.

68

K.L. Brown, L.G. Bacon / Geothermics 38 (2009) 6471

colloids already formed, rather than the formation of new nuclei.

Cooling at higher rates runs the risk of further nucleation and a
bimodal silica sol size distribution. Therefore, the nucleation temperature controls the number of particles that are formed, and the
cooling regime which follows determines the nal particle size.
Thus a low nucleation temperature (20 C) produces a colloid size
of approximately 10 nm while a higher nucleation temperature of
70 C produces a nal particle size of about 60 nm. The use of a
stabiliser (ACUMERTM 5000, from Rohm and Haas) was considered
essential to avoid potential particle agglomeration. This was dosed
at a rate of less than 5 ppm.
3.4. Concentration and stabilisation of the colloid
Commercial silica sols are normally produced at concentrations
ranging from 30% to 50% silica by weight. The challenge for producing similar products from geothermal water involves not only
the ability to control particle size but also to be able to concentrate
sols from a very dilute 300 mg/L to more than 30%, equivalent to a
concentration factor of >1000 times. We chose UF to concentrate
the sol. Since we used a continuous ow plant, a continuous ow
UF system was adopted (Fig. 3). The rst two stages were designed
to concentrate the silica colloid to a few percent and the nal stage
concentrated the solution to >30%. The actual nal concentration
of silica sol was controlled by a ow-rated discharge valve. We
decided on the use of hollow bre cartridge membrane modules for
Stages 1 and 2 and tubular membranes for Stage 3. The use of the
more rugged tubular membranes for the higher silica concentrations in Stage 3 provided for less fouling and easier cleaning of the
membranes. The primary problem with UF techniques is membrane
fouling, which results in declining uxes together with increasing
membrane pressures. Membrane cleaning frequency is therefore
a major economic consideration and the membrane pressure was
continuously monitored to enable performance to be measured and
cleaning intervals to be optimised.
At Wairakei the initial pH of the separated geothermal water is
approximately 8.5 and hence sufciently alkaline to achieve stability of low concentration colloids during the ageing process. The
organic stabiliser used was injected into the ow after ageing and
prior to ultraltration. This was found to be sufcient to inhibit
agglomeration during the process and provide ongoing stability
over several years.

Fig. 4. A normal monodisperse particle size distribution with average particle size
of about 62 nm. Histograms show volume percent for different particle size ranges,
line shows cumulative volume percent.

Some are preferentially adsorbed, such as Ca, Sb and Mg. Future

work could consider the use of chelating agents, for instance, to
reduce the adsorption of multivalent ions onto the colloid.
In general the designed pilot plant performed as anticipated and
operated successfully on an intermittent basis for 11 months. However, it was found necessary to add two large ageing tanks to ensure
complete polymerisation prior to UF. In common with observations
at Wairakei and other NZ geothermal elds, the silica scaling rate is
maximised when both polymeric silica and available monomeric
silica is present simultaneously at elevated temperatures. Initial
cleaning frequency was typically 6 h. This was deemed unsatisfactory; however, installation of the extra ageing tanks decreased
this frequency to about 57 days. The main fouling occurred in the
hollow bre membranes in Stages 1 and 2. A cleaning cycle utilised
hot hydrouoric acid at pH 3 over a period of 12 h.
During the entire operation of the pilot plant, there was very
little silica scaling observed in any of the ageing tanks or on the
permeate side of the ultralters. This is consistent with eld observations in the general power plant indicating that scaling rates are
small at low temperatures when the silica is fully polymerised.

3.5. Dialtration
The dialtration or washing of unwanted salts from the concentrated geothermal silica sols was carried out using local potable
water. The process could be used at any stage during the UF process
though less fresh water, and hence less time, is required at higher sol
concentrations. The inclusion of salts in the colloid matrix during
particle growth was an inevitable consequence of using geothermal
brine. This was ameliorated to some extent at Wairakei where low
dissolved solids concentrations exist.
3.6. Results
Particle size was measured by dynamic light scattering (or photon correlation spectroscopy; Pecora, 1985). Average particle sizes
from about 10 to 70 nm could be reproducibly manufactured. The
size distribution was acceptably monodisperse; a characteristic
example is shown in Fig. 4.
A typical analysis, measured by ICP-MS and ICP-AES, is presented
in Table 2 together with the original geothermal brine concentration. Many of these constituents have little effect on the nal use
of the product. As can be seen from the table, the sol concentrates
most of the ions in solution due to the surface charge on the colloid.

Table 2
Comparison of initial and nal solutions in the silica pilot plant at Wairakei.
Parameter

Raw geothermal uid (ppm)

Produced silica sol (ppm)

SiO2
Cl
Na
K
Li
Ca
SO4
B
F
Br
Rb
Cs
As
Sb
Al
Mg
Fe
pH
Density (g/ml)
Particle size (nm)

490
1688
960
144
9
15
34
22
12
4.7
1.8
2
4.5
0.1
0.3
0.005
<0.1
8.7
n.a.
n.a.

250,000
n.a.
1,960
900
10
1,050
n.a.
100
n.a.
n.a.
20
40
3
80
550
10
3
8.4
1.32
48

n.a.: not analysed.

K.L. Brown, L.G. Bacon / Geothermics 38 (2009) 6471

69

Fig. 6. Silica concentrations in steamlines at Wairakei as a function of pipeline

length.

Fig. 5. Condensate erosioncorrosion in steamline L at Wairakei.

3.7. Conclusions
The experiments proved that it is possible to produce silica sols
that compare favourably with commercial products. One of the
prime requirements of a commercial process is the ability to produce sols of predetermined particle size, which we were able to
accomplish. The purity of the sols is satisfactory, but probably better
results could be achieved with greater time for the dialtration.
The economics of a full-scale process were not easy to ascertain;
however, the principal factors were recognised. Probably the largest
operational cost is the replacement of the ultralters. These need
replacing when they block, and cannot be satisfactorily cleaned.
This is a gradual process, and we never reached this stage in our
experiments. Consequently, we can only estimate the likely lifetime of the lters. Added chemicals are not expensive. The UF units
are expensive, and a plant with any reasonable throughput would
require large areas of land for the ageing tanks. The remainder of
the equipment is not excessively expensive.
The removal of silica from the brine allows a large range of
downstream processes. Other mineral extraction processes become
much easier when the silica is removed and heat exchangers can
also be much more readily utilised without the threat of silica deposition.
4. Steamline corrosion and steamline scrubbing
The Wairakei Power Station was commissioned in 1958 using
steam at a pressure of 1.5 MPa (15 barg) from the Eastern Boreeld, situated about 1.5 km from the A Station turbine-generators.
At that time, steam was supplied via ve, 20-in. diameter (508 mm)
pipelines. The second stage of commissioning saw three additional, 36-inch diameter (762 mm) pipelines installed supplying
additional steam to three, 30-MW turbine-generators in the B Station. Much of this steam was supplied from the Western Boreeld,
situated about 3 km to the west of the Wairakei Power Station.
Steamline condensate corrosion (Fig. 5) discovered in the late
1970s initiated an intensive investigation into the causes of patch
type erosioncorrosion that showed rates of up to 0.3 mm/year
and patch depths of up to 8 mm. It was found that galvanic corrosion was the primary corrosion driver, with the owing condensate
removing iron from the upstream edge of each patch and depositing
magnetite at the downstream edge. As a consequence, condensate
iron levels (measured at condensate drains) remained very low
(1020 ppb) and patch migration occurred in the intervening period
between inspections. The magnetite formed was coarse-grained,
non-protective and easily aked off as a result of pipeline contraction during cooling.

The Wairakei plant has been fortunate with respect to steam

quality, with no signicant turbine deposition or chloride stress
corrosion cracking. Because it was the rst plant in the world to
use steam from a liquid-dominated geothermal system there was a
gradual and progressive increase in generation over a long period
with the installation of steamlines of different diameters feeding
a range of turbines varying in pressure and generation capacity.
By 1975 there were 10 installed pipelines, up to about 3 km in
length, with diameters of 280, 520, 760 and 1200 mm. This provided
the opportunity to investigate steamline scrubbing under various
ambient conditions in pipelines with different types of insulation,
and also a range of drainpot types and efciencies.
The result of these investigations was a numerical scrubbing
model, which assumed a homogenous two-phase mixture of steam
and water droplets at the separator outlet, and calculated steam
and condensate contaminant chemistry (i.e. silica or chloride) at
any point downstream. A scrubbing efciency was determined
for a 760-mm pipeline using measured drainpot silica concentrations and calculated pipeline condensation rates. The scrubbing
efciency of the pipeline was dened as the percentage of contaminant removed from the steam phase into the condensate phase,
and this was made on a per unit length basis to allow for varying
pipeline lengths. To extend the model to other pipeline sizes and
other conditions it was necessary to decide what factors inuence
the scrubbing efciencyvolume ow seemed intuitively to be the
most likely. Scrubbing efciencies between 0.079% and 2.717% per
metre were determined for the pipeline sizes mentioned earlier.
Later development of the model added thermodynamic data to
provide for the prospect of higher steam ows, drier pipelines and
superheat generation. The latter was a new issue for Wairakei which
arose, rstly from the installation of large diameter pipelines with
increasingly efcient insulation, and secondly from the use of highow wells that produced dry steam or (in some cases) superheated
steam.
Fig. 6 provides a comparison of steam silica levels in three
pipelines of different diameters (520 mm, 760 mm, and 1220 mm)
at average steam velocities, with the same insulation properties.
Steam was supplied from a separator with a separation efciency of
99.99%. These graphs demonstrate the decline in scrubbing performance as pipeline diameter increases. A steam silica concentration
of 0.01 ppm in a 520-mm diameter pipeline requires a pipeline
length of approximately 200 m, whereas the same level is achieved
in a 1200-mm pipeline at about 2500 m.
5. Hydrogen sulphide bioreactor
The Wairakei Power Station is the only geothermal plant in the
world with once through cooling. It uses approximately 17 m3 /s
of water from the adjacent Waikato River in direct contact condensers, and discharges both cooling water and condensate to the
river via concrete culverts about 150 m in length. This discharge is

70

K.L. Brown, L.G. Bacon / Geothermics 38 (2009) 6471

Fig. 7. Pipe from the bioreactor pilot plant at Wairakei with adherent biolm of
sulfur oxidizing bacteria (white).

about 15 C hotter than the upstream river water, and is of lower pH

due to carbon dioxide absorption and a hydrogen sulphide (H2 S)
concentration of about 1 mg/L. Shortly after commissioning of the
plant it was discovered that headspaces in the culverts were suffering from severe acid attack due to the presence of H2 S. Under these
conditions, a surface moisture pH of <2 is common in the presence
of sulfur oxidising bacteria which oxidise H2 S to sulphuric
acid.
Investigations, required for re-permitting the plant under the
Resource Management Act of 1991 (Daysh and Chrisp, 2009), into
the effects of the discharge, showed that dissolved H2 S levels in
the Waikato River, after mixing of river water and discharge, were
well in excess of the water quality guideline of 2 ppb. Plant reengineering options that were considered to overcome this problem
included new surface condensers to replace existing direct contact
condensers, closed circuit cooling, separate gas cooler treatment,
helper cooling towers and, lastly, treatment to remove H2 S.
Although the treatment of 17 m3 /s of water was a formidable
task, the previously considered nuisance sulfur bacteria were chosen as a potentially cost-effective treatment. Therefore, a pilot plant
was constructed to demonstrate process viability and to optimize
process parameters. A number of biolm support structures were
investigated, including ringlace ropes and polycarbonate sheets.
The most successful structure was a 100-mm diameter pipe, which
provided complete oxidation of the dissolved H2 S within approximately 120 s. The white biolm of sulfur oxidizing bacteria is shown
in Fig. 7. Typical results from the bioreactor pilot plant (i.e. percent H2 S removal versus distance) are shown in Fig. 8. A provisional
design for a full scale plant based on these results was then undertaken.
6. Arsenic removal from geothermal brine
Separated geothermal water from the Wairakei geothermal
eld containing about 4 mg/L of arsenic (Glover and Mroczek,
2009) has been discharged into the adjacent Waikato River since
the commissioning of the power station in 1958. In the mid1990s investigations into the feasibility and economics of removing
the arsenic were directed at several processes, including arsenic
sulphide precipitation using the hydrogen sulphide in station offgases.
An arsenic removal pilot plant was set up at Wairakei, using
the process shown in Fig. 9. The station off-gasses were fed into
the base of the gas absorption tower where they contacted a
counter current ow of geothermal water. A small ow of gassaturated water was fed forward from this process to mix with

Fig. 8. Typical results from the pipe bioreactor at Wairakei.

Fig. 9. Schematic diagram of the arsenic removal plant at Wairakei.

the 40 m3 /h of geothermal water to be treated. Separation of solids

was achieved using dissolved air otation following the addition of
occulants.
The experimental data indicated that: (a) At optimum gas and
water ows the gas absorption tower was only able to dissolve H2 S
to about 30% saturation; (b) A brine pH of approximately 5 was
required for maximum orpiment (arsenic sulphide) precipitation;
(c) Of the large number of occulants tested only two succeeded in
aggregating orpiment particles. These were Buckman Laboratories
BULOC 535 (a bridging polymer) and BL5117 (an anionic polyacrylimide). Both were required at a concentration of 20 ppm each;
(d) Arsenic removal at optimum pH was approximately 40%; (e) An
analysis of the DAF tank oat showed arsenic and sulfur contents
of 948 and 700 mg/kg respectively, suggesting that some colloidal
sulfur was included; and (f) Potential environmental effects (with
respect to river discharge) include the presence of excess dissolved
H2 S and possibly residual orpiment and sulfur unless complete separation could be achieved.
Finally, it was concluded that the proposed arsenic removal process, despite its simplicity, would not be economic due to the high
cost of occulants.

K.L. Brown, L.G. Bacon / Geothermics 38 (2009) 6471

7. Summary of the pilot plant experiments

The results of the pilot plant tests showed that:
1. The effects of ow on silica scaling could be investigated and
characterized by simultaneously controlling the size of the colloidal silica particles and uid ow conditions, and observing the
silica deposition on at plates and vertical cylinders.
2. By-product silica sols could be prepared by using temperature
control to regulate sol particle size, followed by ultraltration to
concentrate the sol and dialtration to remove contaminants.
3. Sulfur oxidising bacteria efciently remove dissolved hydrogen
sulphide from water discharges using pipelines as biolm support structures.
4. Arsenic solids can be successfully separated using dissolved air
otation following the addition of H2 S and occulants.
5. Galvanic corrosion is the primary driver in patch type
erosioncorrosion in steam transmission lines, with owing condensate removing iron from the upstream edge of each patch and
magnetite deposition at the downstream edge.
6. Steam scrubbing performance decreased as steam-line diameter
increased.
Acknowledgements
We thank Ian Thain, who, as station manager of Wairakei Power
Station, actively supported these pilot plant trials. Dr. Mike Dunstall
and Wayne Christie co-operated in some of the trials.

71

References
Brown, K.L., Bacon, L.G., 2000. Manufacture of silica sols from separated geothermal
water. In: Proceedings of the 2000 World Geothermal Congress. Kyushu-Tohoku,
Japan, pp. 534537.
Brown, K.L., Dunstall, M., 2000. Silica scaling under controlled hydrodynamic conditions. In: Proceedings of the 2000 World Geothermal Congress. Kyushu-Tohoku,
Japan, pp. 30393044.
Daysh, S., Chrisp, M., 2009. Environmental planning and consenting for Wairakei:
19532008. Geothermics 38, 192199.
Dunstall, M., Brown, K.L., 1998. Silica scaling under controlled hydrodynamic conditions: vertical at plate and vertical cylinder test. In: Proceedings of the 23rd
Workshop on Geothermal Reservoir Engineering. Stanford University, Stanford,
CA, USA, pp. 204210.
Dunstall, M., Zipfel, H., Brown, K.L., 2000. The onset of silica scaling around circular
cylinders. In: Proceedings of the 2000 World Geothermal Congress. KyushuTohoku, Japan, pp. 30453050.
Garibaldi, F., 1980. The Effect of Some Hydrodynamic Parameters on Silica Deposition. Geothermal Institute, University of Auckland, New Zealand, Diploma
Project 80.11, 44 pp.
Gunnarsson, I., Arnrsson, S., 2005. Impact of silica scaling on the efciency of
heat extraction from high-temperature geothermal uids. Geothermics 34,
320329.
Glover, R.B., Mroczek, E.K., 2009. Chemical changes in natural features and well discharges in response to production at Wairakei, Geothermics, New Zealand, 38,
117133.
Pecora, R., 1985. Dynamic Light Scattering: Applications of Photon Correlation Spectroscopy. Plenum Press, New York, NY, USA, 436 pp.
Pott, J., Dunstall, M.G., Brown, K.L., 1996. Numerical simulation of silica scaling.
In: Proceedings of the 18th New Zealand Geothermal Workshop. University of
Auckland, Auckland, New Zealand, pp. 4146.
Weres, O., Yee, A., Tsao, L., 1982. Kinetic equations and type curves for predicting the
polymerization of amorphous silica in geothermal brines. Society of Petroleum
Engineers Journal 22 (1), 916.