Professional Documents
Culture Documents
Geothermics
journal homepage: www.elsevier.com/locate/geothermics
a r t i c l e i n f o a b s t r a c t
Article history: The organic rankine cycle (ORC) is a heat recovery technology used for renewable energy generation such
Received 5 August 2015 as the large-scale production of electricity from geothermal resources. In this paper a dynamic model of
Received in revised form a commercial-scale geothermal ORC is developed in process simulation software and validated against
30 November 2015
plant data. The difference in power output between the model and plant was found to be 0.24% with a
Accepted 10 January 2016
standard deviation of 1.40% of the average power output. The results from model validation are used to
Available online 29 January 2016
suggest improvements to the model, which is intended to be used for further investigation of optimisation
and process control.
Keywords:
Organic rankine cycle © 2016 Elsevier Ltd. All rights reserved.
ORC
Dynamic modelling
Model validation
1. Introduction Large scale geothermal ORCs are qualitatively different from the
small scale plants because of the larger number of interconnected
In this paper, a dynamic model of a commercial scale geother- unit operations in these processes. Due to the increased complexity,
mal organic rankine cycle (ORC) plant in Northland, New Zealand is there is a greater possibility of unanticipated consequences due to
presented. The model is constructed in a process flowsheet simula- interactions between components. For this reason, a process flow-
tor, VMGSim, using design data, and is then validated with 24 h of sheet simulator was used to build the model. Process flowsheet
plant data. The output of the model is compared to the behaviour simulation software is useful for modelling large scale processes
of the real plant. From these results, we assess the effectiveness compared with equation based or ad hoc models developed specifi-
of process flowsheet simulators in the modelling of these types of cally for a particular process, because the individual unit operations
plants, as well as identify particular parts of the model where its are modelled separately, which allows a process to be modelled
accuracy could be improved. using modular “building blocks”. This makes it easier to capture the
Dynamic models of processes are useful for understanding potentially complex behaviour of large scale processes compared
transient behaviour such as process stability, the effectiveness of with a model that is specific to one process.
different control philosophies, startup and shutdown, and emer- If process flowsheet simulators are to be used to create dynamic
gency situations like turbine trips and lifting of safety valves. The models of geothermal ORCs in the future, we must first verify
model described in this paper focusses on normal operation and its their effectiveness in modelling such plants. The novelty of this
usefulness is thus limited to process stability and analysis of control paper lies in its dynamic modelling of a large (on the order of tens
systems. This is due to the lack of plant data and design information of megawatts electric power production) commercial geothermal
needed to validate the more uncommon modes of behaviour. How- ORC power plant, and also the use of a process flowsheet simulator
ever, due to the modular nature of process flowsheet simulators to build this model. The results of this work will be of interest to
this model serves as a good first step to build a more sophisticated the designers and operators of these plants as it provides confidence
model, should this information become available. that these tools can be used to investigate process operability. In
addition, as discussed in Section 3.2.4 of the paper, the novel use of
a turbomachinery performance curve in the model is more likely
to more accurately capture the behaviour of a fixed speed turbine
compared with some of the approaches used in the literature.
∗ Corresponding author.
E-mail address: w.yu@auckland.ac.nz (W. Yu).
http://dx.doi.org/10.1016/j.geothermics.2016.01.007
0375-6505/© 2016 Elsevier Ltd. All rights reserved.
64 M.J. Proctor et al. / Geothermics 61 (2016) 63–74
and NCG preheater, respectively. The streams then recombine (S8) 3.2. Model construction
before passing through the shell of the vaporiser, where the fluid
becomes a saturated vapour. The fluid is then split into two streams To create a model with VMGSim, connections between unit
that pass through two identical expanders (again, only one stream operations, equipment specifications and stream characteristics
is shown). The outlets of the expanders pass through the recu- are specified. The connections between unit operations reflect the
perators before being recombined and sent through the air cooled layout of the plant as shown in Fig. 2. We provided the thermo-
condenser. Here the vapour is condensed and returned to the pump dynamic state, flow rate and composition of the geothermal input
to complete the cycle. stream and cooling air, as well as the composition of the working
fluid (n-pentane). We also specified the pressure at every stream
on the boundary of the model. Details of particular unit operations
were also provided to the simulator. Constant UA values (the overall
heat transfer coefficient, U, multiplied by the heat exchanger area,
A) were provided for each heat exchanger based on plant design
3. Method data and the pump curve for the manufacturer was imported into
VMGSim. The volume of each unit operation was also found from
3.1. Simulation software the plant technical drawings and included in the model. Due to the
commercial sensitivity of this information precise values of these
There are a wide variety of process flowsheet simulators such specifications are not provided in the paper. Some approximate
as Aspen HYSYS, Modelica, VMGSim and Aspen Plus, which are specifications are given in Table 1. It should be noted the UA val-
used by both industry and academia to analyse the steady state ues shown for the vaporiser and preheater are subsequently scaled
and dynamic behaviours of chemical processes. VMGSim (v8.0) was down according to the fine-tuning process described in Section 3.5.
selected to build the model in this paper. The sections of the control system essential to regulating the
VMGSim was originally developed by Virtual Materials Group plant were also included. This consisted of the level controller for
(Cota et al., 2003). It has been used around the world to model the geothermal separators, the level controller for the shell of the
existing processes and to design new facilities. Its thermo- vaporizer and working fluid temperature control for the preheaters.
physical property calculator is capable of accurately predicting the Control valve flow coefficient (Cv) values provided by the plant
behaviour of complex fluid mixtures. Previously, members of our operator were also entered into the model.
research group used VMGSim in a geothermal application to evalu-
ate the enhancement of the flash-cycle Kawerau geothermal power
plant in New Zealand (Dabbouret al., 2011).
66 M.J. Proctor et al. / Geothermics 61 (2016) 63–74
Table 1 exchanger and so the calculation used to find the heat duties of the
Approximate values of the specifications used in the model.
coolers is log mean temperature difference (LMTD), considering
Specification Value each set of tubes separately, as shown in Eq. (1) below. As men-
Turbine design power 10 MW tioned in Section 3.2, UA values were calculated from plant design
Turbine inlet design flow 5000 m3 /h data and assumed to be constant.
Turbine speed 3000 RPM (fixed, as turbine is
coupled to the electrical grid) Q̇ = UA × LMTD (1)
Turbine design efficiency 70%
Recuperator UA 300 kW/K where:
Condenser UA 7000 kW/K
Preheater UA 600 kW/K
Th,in − Tc,out − Th,out − Tc,in
LMTD = (2)
Brine preheater UA 2000 kW/K (Th,in −Tc,out )
Vaporiser steam UA 2000 kW/K
ln
(Th,out −Tc,in )
Vaporiser brine UA 800 kW/K
NCG preheater UA 150 kW/K where the subscripts ‘h’ and ‘c’ stand for the hot and cold stream
Geothermal reinjection pressure 1000 kPa respectively (with the hot stream being either the geothermal
Ambient pressure 101.33 kPa steam or brine and the cold stream being the working fluid). The
Preheater shell volume 10 m3
Preheater tube volume 1 m3
subscripts ‘in’ and ‘out’ represent the inlet or outlet of the heat
Brine preheater shell volume 20 m3 exchanger respectively and Q̇ refers to the amount of heat trans-
Brine preheater tube volume 5 m3 ferred per unit time.
Vaporiser shell volume 30 m3 As can be seen in Fig. 3, an iterative process is used to model the
Vaporiser steam tube volume 1 m3
vaporiser. An input/output diagram of the vaporiser sub-model is
Vaporiser brine tube volume 1 m3
Recuperator shell volume 10 m3 given in Fig. 5 below. The inputs of the sub-model are the temper-
Recuperator tube volume 3 m3 atures of the steam, brine and working fluid inlet streams, and the
Air condenser volume 10 m3 mass flow rates and heat capacities of these streams. The outputs
are the temperatures of each outlet stream. At each time step, the
heat transferred between the streams is calculated. The inputs are
3.2.1. Vaporiser updated at each time step by other unit operations in the model
A heat exchanger with multiple tube sets is not directly pro-
vided by VMGSim. To model the vaporiser we constructed a lumped
parameter model from simpler unit operations already available in
the simulator. Fig. 3 shows a flow diagram which explains how the
vaporiser calculations are performed and Fig. 4 shows the true lay-
out of the vaporiser as it exists in the plant, including the location
of the temperature variables T1 to T6 that exist in Fig. 3.
In order to model the vaporiser (where the shell contains two
separate sets of tubes carrying the geothermal brine and steam)
the heat exchanger was split into three unit operations: two cool-
ers representing the tubes, and a heater representing the shell. The
heat duties of the coolers are summed and provided to the heater
in order to specify its heating duty while maintaining conserva-
tion of energy. The vaporiser is modelled as a counter current heat Fig. 4. A diagram of the true vaporiser layout.
M.J. Proctor et al. / Geothermics 61 (2016) 63–74 67
to be calculated. This would allow the model to predict the amount Table 2
The composition of the geothermal fluid.
of subcooling.
Chemical species Mole fraction
3.2.4. Control system Water 0.99527
The control system was modelled by including the PI Carbon dioxide 0.00307
(Proportional-Integral) control loops in the plant, along with the Nitrogen 0.00166
controller gain and integral time settings as advised by the plant
operator. The loops which have been modelled are:
12.9 194
• Vaporiser working fluid liquid level 12.8
• Main geothermal separator liquid level 193.5
12.7
• Vaporiser condensate liquid level
Temperature (C)
Pressure (bar)
• Net power output, and 12.6 193
• Vaporiser geothermal fluid pressure
12.5
192.5
In total, there are eighteen separate control loops used in the 12.4
plant. However, most of these have been excluded as they are
12.3
outside the scope of the model. This is because they deal with condi- 192
tions outside normal operation such as start-up, shutdown or safety 12.2
systems, such as the turbine or vaporiser bypass lines. Similarly, 12.1 191.5
the bypass lines that serve a safety function in the plant were not 0 2 4 6 8 10 12 14 16 18 20 22 24
included in the model. This was because of the lack of plant data
Time (hours)
required to validate these sections of the plant as they will only be
active under emergency situations which did not occur during the
Geothermal Fluid Pressure
period of time for which data was available for this study.
Geothermal Fluid Temperature
3.2.5. Geothermal fluid
Fig. 6. Real plant geothermal fluid inlet conditions, contrasting geothermal fluid
The geothermal fluid is a complicated mixture of geothermal
pressure with temperature.
brine and steam containing NCGs. In this section the properties
of the fluids used in the model, as well as the composition of the
geothermal fluid are discussed. could be re-run in an attempt to obtain more accurate results. Also,
The simulator uses a built-in thermodynamic database of fluid more chemical species could be included in the analysis based on a
properties. All of the fluids, or components of those fluids used in greater understanding of the chemistry of the geothermal reservoir.
the model were found in this database. The components used for
the geothermal fluid were water, carbon dioxide and nitrogen, the 3.2.6. Self-consistency of geothermal inlet conditions
working fluid was pure n-pentane and the cooling air was specified Examination of the plant data for the geothermal fluid pressure
using the aggregate properties of air (present as an option in the and temperature at the inlet revealed a problem. The temperature
thermodynamic database) directly. given was too high, and would produce a stream that was entirely
This was used in conjunction with a thermodynamic model vapour, instead of a two phase mixture as expected.
called the advanced Peng Robinson equation of state (EOS) to pre- This is caused by the temperature sensor being present in the
dict fluid thermodynamic properties. The advanced Peng Robinson liquid stream exiting the main geothermal separator. Due to the
EOS is an extension of the Peng Robinson EOS that uses a more composition difference between the liquid (mostly water) and
refined asymmetric mixing rule for multicomponent fluid equilib- vapour (a mixture of water and noncondensable gases) and a lack
ria. The advantage of using this method is more accurate liquid of thermodynamic equilibrium in the separator, the temperature of
density estimation compared with the regular Peng Robinson EOS, the liquid is slightly higher than that of the vapour. However, the
particularly for polar compounds (Mathias et al., 1991). The Peng simulator assumes that the separator is at an equilibrium state.
Robinson EOS was developed as an improvement over prior equa- To get around this problem, the vapour fraction of the geother-
tions of state which are all based ultimately on the van der Waals mal fluid inlet stream was assumed constant, and the temperature
equation. It is described in (Peng and Robinson, 1976) as being par- set to vary to achieve that vapour fraction regardless of the
ticularly good compared with its predecessors in estimating liquid geothermal pressure. This assumption ties the geothermal fluid
density values. temperature to the pressure, and the temperature will rise and fall
The composition of the geothermal fluid was found through a along with the pressure.
least squares optimisation. Several chemical species known to be As can be seen in Fig. 6, which plots the real plant data for
in the geothermal fluid were specified in a steady state model of geothermal fluid and temperature, this is not an unreasonable
the plant. The mole fractions of these components were specified assumption as these variables are well correlated, with a correlation
at the plant inlet and then were varied to minimise an error func- coefficient of 0.97.
tion consisting of the differences between the model and design
data for total enthalpies at the outlets of the heat exchangers. This 3.2.7. Model fine-tuning
found a composition, shown in Table 2, which minimised the error After the model was built and tested based on design data, the
in enthalpy across the plant. UA parameters of the vaporiser and preheaters (for which approx-
Other species tried were sulphur dioxide, sulphur trioxide and imate values are shown in Table 1) were tuned in an attempt to
hydrogen sulphide, however it was found that these components improve the power output prediction of the model.
minimised the error function when set to zero. This optimisation This was done by first preserving the ratios between the UA
is dependent on the accuracy of the design data, which may only parameters, then scaling down the vaporiser steam tube UA by
be an indication of the real conditions throughout the plant. In the multiplying it by a fixed value of 0.84, thereby lowering all of the
future, when more data from the plant is obtained the optimisation heat exchanger UA values. This value was chosen based on a least
M.J. Proctor et al. / Geothermics 61 (2016) 63–74 69
Power (MW)
17
16
3.3. Model validation
15
The model was validated by using 24 h of recorded data from 0 2 4 6 8 10 12 14 16 18 20 22 24
the plant’s control system as input to the model and comparing the Time (hours)
model output to the behaviour of the real plant. Tuned Model Power Output
The five inputs to the model are: Plant Power Output
Untuned Model Power Output
• Geothermal fluid pressure Fig. 7. Gross power output comparison plot. (For interpretation of the references to
• Geothermal fluid temperature color in the text, the reader is referred to the web version of this article.)
• Ambient temperature
• Vaporiser level controller set point
• NCG vent mass flow rate set point trol system design, which depend on the internal state of the plant
being modelled accurately for any results to be useful.
12.8 14 1.5
12.7 13
1.4
12
12.6
Pressure (bar)
Pressure (bar)
11
12.5 1.3
10
12.4
9 1.2
12.3
8
1.1
12.2 7
12.1 6 1
0 2 4 6 8 10 12 14 16 18 20 22 24 0 2 4 6 8 10 12 14 16 18 20 22 24
Time (hours) Time (hours)
The use of the built-in performance curve for the turbine could
35
also cause inaccuracy by incorrectly predicting the isentropic effi-
ciency and pressure-flow resistance of the turbine, but based on
the discussion in this section we believe that most of the model
30
mismatch is caused by the three factors mentioned above, and that
Temperature (C)
the impact of differences between the true performance curve and
modelled performance curve is minimal. 25
Based on looking at these three variables we conclude that the
model mismatch in gross power output, mainly existing between
ten and eighteen hours, is caused by inaccurate modelling of the 20
vaporiser and air cooled condenser. The vaporiser also accounts for
the smaller amount of mismatch present between zero and four
hours. 15
Table 3 shows summarising statistics of the residual for all the 0 2 4 6 8 10 12 14 16 18 20 22 24
variables tracked in the model. These are the absolute value of the Time (hours)
average and standard deviation, as well as a percentage value with
respect to the average of the real plant data. The percentage values Actual Condenser Outlet T
are of most interest, since they are more indicative of model inaccu-
racy than the absolute values. For example, the standard deviation Model Condenser Outlet T
for the brine mass flow rate is 9.77 t/h, which is high in comparison
Fig. 10. Condenser outlet temperature comparison plot.
to the others, but its percentage value with respect to the average
real brine flow is 1.81%, which puts this into perspective.
Looking through the variables in Table 3 it can be seen that most The turbine outlet pressure is determined by two factors. The
of the average and standard deviation values are small. This means first is adequate modelling of the condenser temperature. The
the residuals are close to the real output and stay reasonably close to working fluid must be at vapour–liquid equilibrium as it condenses,
their averages, which means that the offsets are small and reason- so the condenser pressure will be dependent on the temperature.
ably constant over a range of model inputs. This is exemplified by The second is the resistance to flow in the recuperator and con-
the power output of the plant, which is the most important output denser, which is dependent of the geometry of these unit operations
variable from the model. However, not all variables are predicted as well as the flow rate through the plant.
accurately. For example, the results for the condenser show the As seen in Fig. 10, the condenser outlet temperature increases
outlet pressure and temperature are quite large. This indicates an during the day due to the increase in ambient temperature. It can be
area where the model can be improved. seen that the change in condenser temperature is higher in the real
plant compared to the model. This is why the turbine pressure in
4.1. Impact of air cooled condenser modelling the real plant increases more than is predicted during the day as the
ambient temperature rises, which in turn causes most of the model
Fig. 9 shows the real and modelled turbine outlet pressure. It mismatch in gross power output. The flow resistance impacts the
is apparent that there is an offset between the real and predicted model mismatch to a lesser degree, and can be refined by selecting
results. This offset is not relevant for our analysis as it does not have an appropriate ‘k’ value for the resistance, as per Eq. (3).
a substantial impact on the gross power residual; the important The accuracy of the condenser outlet temperature is determined
thing to note is the change in the pressure between night and day by the condenser model. As mentioned in Section 3.2.3, the con-
is greater for the real pressure compared with the model pressure. denser is modelled with a constant UA value and subcooling is not
An increase in turbine outlet pressure will result in a decrease in present. It may be possible to improve the accuracy of the con-
the gross power output. This contributes to the real gross power denser outlet temperature by using heat transfer correlations to
output having a larger decrease between night and day than the model how the UA value will change with different flow condi-
model. tions. This would take into account both the flow conditions inside
M.J. Proctor et al. / Geothermics 61 (2016) 63–74 71
Table 3
Summarising statistics for the residuals (model result minus real data) of all variables tracked in the model.
Unit operation Fluid Stream Measurement Average of the residual Standard deviation
(unit) of the residual
19.5 Figure where there is a drop in the real power output at around
four hours as a consequence of the drop in vaporiser pressure. The
vaporiser shell pressure prediction also has a major deviation from
19 the real data between twelve and nineteen hours. This will also con-
tribute to the inaccuracy of the power output prediction, but as it
Pressure (bar)
is coincident with the effect of the modelling of the air cooled con-
denser (as mentioned in Section 4.1), the difference between the
18.5
curves in Fig. 7 cannot be exclusively assigned to one effect or the
other. Finally, there is an interesting feature present in the model
curve starting at nine hours where it rises by around five kilopas-
18 cals; the real data appears to remain relatively constant during this
period.
The small increase in the modelled vaporiser shell pressure at
17.5 around nine hours appears to be solely related to an increase in the
0 2 4 6 8 10 12 14 16 18 20 22 24 ambient temperature at this time. In the model, when the ambi-
Time (hours) ent temperature is increased this increases the temperature of the
working fluid exiting the condenser. Although the amount of heat
Actual Vap Shell P Model Vap Shell P transfer into the working fluid decreases when this happens, this
effect does not outweigh the increase in working fluid temperature,
Fig. 11. Vaporiser shell pressure comparison plot. resulting in a higher temperature at the outlet of the vaporiser, and
subsequently a higher pressure. The effect of this is not very large,
the tubes for the working fluid, for example accounting for the two- as seen by the comparatively small change in the vaporiser pressure
phase flow and potential subcooling, as well as outside the tubes during this period, but due to the lack of a corresponding increase
for the air flow. Currently the air flow rate is modelled using Eq. in the real vaporiser pressure this must be considered an artefact
(3) with a fixed pressure drop and flow resistance. Modelling the of the model.
performance of the fans with varying ambient temperature and The largest deviation occurs between twelve and nineteen
pressure could also improve accuracy by getting a better estimate hours. In the real plant the vaporiser pressure reduced by about
of air flow through the condenser. 0.3 bar, but the model only decreased by 0.1 bar. This is related to
the overall variability of the real data compared with the model,
where the model has far less variation. The main driver of the drop
4.2. Impact of vaporiser modelling in vaporiser pressure in this case appears to be the geothermal flow,
which also decreases during this period. What is different between
The vaporiser shell pressure influences the pressure drop the model and real data is the degree to which they respond to the
through the turbine and also provides an indication of the work- reduction in geothermal flow rate. One possible cause is the higher
ing fluid mass flow rate. This is because the resistance to flow temperature of geothermal fluid exiting the vaporiser in the model
through the turbine will increase with an increased working fluid compared with the real plant (for example, as shown in Table 3 the
flow rate, resulting in a higher vaporiser pressure. There is good geothermal steam exiting the vaporiser in the model has a tem-
overall agreement between the model and real vaporiser pressures, perature 8.29 ◦ C higher than the real data). The geothermal fluid
with a deviation of 0.27% as shown in Table 3. However, Fig. 11 temperature can decrease more as it passes through the vaporiser
shows that there is some discrepancy in the shape of the curves without lowering the LMTD in the vaporiser as much as would occur
produced by the model and the real data. Also, the real vaporiser in the real plant, resulting in comparatively higher heat transfer for
shell pressure has a larger range than that predicted by the model, the model. The effect of this is a smaller change in vaporiser shell
which indicates the real working fluid mass flow rate varied more pressure when the geothermal flow rate is reduced.
than predicted. For the discrepancy at around zero to four hours, both the
The first area of discrepancy exists between zero and four hours. ambient temperature and mass flow rate of geothermal fluid are
The effect of this on the gross power output can be observed in
72 M.J. Proctor et al. / Geothermics 61 (2016) 63–74
175 62 19
60 18.8
170
Temperature (C)
58 18.6
Pressure (bar)
165
56 18.4
160 54 18.2
52 18
155
0 2 4 6 8 10 12 14 16 18 20 22 24 50 17.8
180
195
210
225
240
255
270
285
300
315
330
345
360
375
Time (hours)
Time (minutes)
Actual Vap Brine Outlet T
Model Vap Brine Outlet T Vap Shell Liquid Level Vap Shell P
Fig. 12. Vaporiser brine tubes outlet temperature comparison plot. Fig. 14. Comparison between vaporiser liquid level and vaporiser pressure.
65 165
163
Temperature (C)
60
Liquid Level (%)
161
159
55
157
155
0 2 4 6 8 10 12 14 16 18 20 22 24
50
Time (hours)
0 2 4 6 8 10 12 14 16 18 20 22 24
Time (hours) Actual Vap Shell Outlet T
Model Vap Shell Outlet T
Actual Vap Shell Level Model Vap Shell Level
Fig. 15. Vaporiser shell outlet temperature comparison plot.
Fig. 13. Vaporiser shell liquid level comparison plot.
relatively constant. Therefore the change in evaporation (and thus both superheating as well as a decrease in the heat transfer coeffi-
the heat transfer) that is implied by Fig. 11 is not due to any of the cient. The second is that a decrease in the level could also cause a
known factors. The input brine temperature sensor also does not change in how the fluid circulated within the vaporiser shell, which
show a large change at this time. However, the outlet brine temper- has various baffles installed in an attempt to enhance mixing and
ature from the vaporiser does show a sharp increase in temperature heat transfer, and this change in flow patterns resulted in a decrease
at around four hours, as shown in Fig. 12. As can be seen in this Fig- in the heat transfer coefficient.
ure, the model does not predict the increase in temperature that In order for either of these possibilities to be true it must first be
did happen in the real plant. established that the change in level was the cause of the change in
The mass flow rate of geothermal fluid and its input tempera- heat transfer coefficient, rather than the operator changing the level
ture did not change but the outlet temperature did. This indicates in response to this change. As can be seen in Fig. 14, which shows a
that the heat transfer coefficient suddenly decreased at around four more detailed view of the plant data over this period, the changes
hours, and this is the cause of the initial deviation of the vaporiser in the two variables occur almost instantaneously. This seems to
pressure from the model. Our assumption was that the vaporiser indicate the change in level is the cause of the change in vaporiser
is typically operated with all tubes submerged, so (as mentioned pressure, as the level set point is under manual control it is unlikely
in Section 3.2.1), the UA value of the vaporiser was assumed to be that the operator would have responded so quickly to changes in
constant. However, it appears that this assumption is sufficiently the vaporiser pressure. In addition, if the level does affect the heat
incorrect to cause model mismatch. transfer coefficient then you would expect the vaporiser pressure
It should be noted that the liquid level in the vaporiser shell, to react almost instantly, as occurs in Fig. 14.
shown in Fig. 13, also changes around this time in response to an A further argument in favour of the change in level influenc-
operator changing the set point on the vaporiser level controller. ing the change in heat transfer coefficient for the vaporiser is the
It is possible that this is the cause of the reduction in heat transfer outlet temperature of the vaporiser, which is shown in Fig. 15. If
coefficient that invalidated our assumption. There are two possi- there were no superheating, as would be expected if the tubes con-
bilities. The first is that the level was lowered to the point where taining the geothermal steam and brine were submerged in the
some of the tubes were exposed to pentane vapour, resulting in boiling working fluid, then the outlet temperature of the working
M.J. Proctor et al. / Geothermics 61 (2016) 63–74 73
fluid from the vaporiser should be expected to drop with the vapor- Fine-tuning of the UA values in the preheaters has shown that
iser shell pressure. Yet data from the plant shows it increases, which the accuracy of the power output prediction can be increased. This
shows there must be some superheating. The fact that it increases can be seen visually and can be quantified as the difference in the
around the time we see the heat transfer coefficient and liquid level sum of the squared differences at each time point between the
decrease indicates that the reduction in level caused some tubes model output and real data, which were 347 MW2 and 84.5 MW2 for
to be exposed to pentane vapour, causing all the phenomena dis- the initial model and the fine-tuned model respectively, a reduction
cussed above. This causes the initial mismatch in vaporiser shell of 76%.
pressure, from zero to four hours. Provided that an accurate model for a plant is available it can
As determined by this analysis, the accuracy of the vaporiser be used to test potential improvements to the control system con-
shell pressure, and subsequently the accuracy of the gross power figuration or the addition of new equipment at very low cost prior
output can be improved by modifying the vaporiser model. This to implementing changes physically. It is hoped that these models
was already done in the fine-tuning, but the UA value was sim- will be useful as generic base models for an ORC plant which can
ply scaled by a constant value. While this reduction improved the be extended to approximate particular plants for the purposes of
accuracy of the power output, it also resulted in an increase in the optimization and control system design.
outlet temperature of the geothermal fluid through the vaporiser,
which as discussed in this section caused further discrepancies in Acknowledgments
the vaporiser shell pressure predicted by the model. To some degree
this is a trade-off between the accuracy of different model outputs, This work is part of the Above Ground Geothermal and Allied
but could be improved with more data as this would allow the Technology (AGGAT) Programme supported with funding from the
UA values of the different heat exchangers throughout the plant New Zealand Ministry for Business Innovation and Employment
to be determined more accurately instead of scaling them all by a contract: HERX1201 administered through the Heavy Engineering
common factor. Further improvements are possible if heat transfer Research Association (HERA).
correlations were used to determine the heat transfer coefficients The authors would like to thank Ray Robinson and his team at
for tubes that are submerged in boiling pentane, and coefficients for Ngawha Generation for their advice and for providing data used in
tubes that are exposed to pentane vapour. The total UA value would this analysis.
be a linear combination of these two heat transfer coefficients, with
the contribution from each dependent on the liquid level.
References
Astolfi, M., Xodo, L., Romano, M.C., Macchi, E., 2011. Technical and economical
4.2.1. Analysis of vaporiser shell dynamic response analysis of a solar–geothermal hybrid plant based on an organic rankine cycle.
The vaporiser shell liquid level is dependent on the liquid mass Geothermics 40 (1), 58–68, http://dx.doi.org/10.1016/j.geothermics.2010.09.
flow rate into the vaporiser shell as well as the evaporation rate. It 009.
Bao, J., Zhao, L., 2013. A review of working fluid and expander selections for
is subject to PI control by manipulation of the control valve located organic rankine cycle. Renew. Sustain. Energy Rev. 24 (0), 325–342, http://dx.
after the pumps. Therefore, the model mismatch is zero, apart from doi.org/10.1016/j.rser.2013.03.040.
when the controller set point was changed (or if there were any Casella, F., Mathijssen, T., Colonna, P., van Buijtenen, J., 2013. Dynamic modeling of
organic rankine cycle power systems. J. Eng. Gas Turbines Power 135 (4),
major disturbances, which did not occur during the period under http://dx.doi.org/10.1115/1.4023120, 042310-1-042310-12.
examination). Examining Fig. 13 shows a close agreement between Chen, H., Goswami, D.Y., Stefanakos, E.K., 2010. A review of thermodynamic cycles
the model and real plant outputs, with a reasonable match between and working fluids for the conversion of low-grade heat. Renew. Sustain.
Energy Rev. 14 (9), 3059–3067, http://dx.doi.org/10.1016/j.rser.2010.07.006.
the model and real plant even as the set point is changed. There Cota, R., Satyro, M., Morris, C., Svrcek, W.Y.Y., 2003. Development of an open source
are small differences upon a set point change, but these are of an chemical process simulator. Proceedings of the IASTED International
extremely short duration. Conference on Modelling and Simulation (MS 2003), 525–530.
Dabbour M., Villena J., Kirkpatrick R., Young B.R., Yu W. (2011). Geothermal reboiler
process development modelling. Chemeca 2011, Sydney, Australia. 2369-2382.
Dixon, S.L., Hall, C.A., 2010. Fluid Mechanics and Thermodynamics of
5. Conclusions Turbomachinery, 6th ed. Butterworth-Heinemann/Elsevier, Burlington, MA.
Felgner, F., Exel, L., Frey, G., 2011. Component-oriented ORC plant modeling for
efficient system design and profitability prediction. Clean Electrical Power
It can be concluded based on the above analysis that the model (ICCEP), 2011 International Conference On, 196–203.
has a reasonable amount of accuracy, with an average difference of Ghasemi, H., Paci, M., Tizzanini, A., Mitsos, A., 2012. Modeling and optimization of
0.24% and standard deviation of 1.40% for the power output predic- a binary geothermal power plant. Energy 50, 412–428, http://dx.doi.org/10.
1016/j.energy.2012.10.039.
tion. From this we conclude that process flowsheet simulators are Hettiarachchi, H.D.M., Golubovic, M., Worek, W.M., Ikegami, Y., 2007. Optimum
able to be used for dynamic modelling of geothermal ORC plants. design criteria for an organic rankine cycle using low-temperature geothermal
More plant data is required to fully validate the modelling of pro- heat sources. Energy 32 (9), 1698–1706, http://dx.doi.org/10.1016/j.energy.
2007.01.005.
cess transients, but validation of changes in the vaporiser shell Kanoglu, M., 2002. Exergy analysis of a dual-level binary geothermal power plant.
liquid level indicate that the model is performing adequately, at Geothermics 31 (6), 709–724, http://dx.doi.org/10.1016/S0375-
least for normal operating conditions. 6505(02)00032-9.
Larjola, J., 1995. Electricity from industrial waste heat using high-speed organic
Based on an interpretation of the model validation results we rankine cycle (ORC). Int. J. Prod. Econ. 41 (1–3), 227–235, http://dx.doi.org/10.
believe the main limitations of this approach relate to using con- 1016/0925-5273(94)00098-0.
stant UA values in the heat exchanger models. These could be Liu, B., Chien, K., Wang, C., 2004. Effect of working fluids on organic rankine cycle
for waste heat recovery. Energy 29 (8), 1207–1217, http://dx.doi.org/10.1016/j.
improved primarily through more sophisticated modelling of the energy.2004.01.004.
heat transfer coefficient. By using heat transfer correlations, the Mathias, P.M., Klotz, H.C., Prausnitz, J.M., 1991. Equation-of-state mixing rules for
effect of changing conditions on the heat transfer coefficient in multicomponent mixtures: the problem of invariance. Fluid Phase Equilib. 67
(0), 31–44, http://dx.doi.org/10.1016/0378-3812(91) 90045-9.
the vaporiser and condenser in particular may be more accurately
Nguyen, T.Q., Slawnwhite, J.D., Boulama, K.G., 2010. Power generation from
modelled, which should reduce the variation in the residual power residual industrial heat. Energy Convers. Manage. 51 (11), 2220–2229, http://
output. For the condenser, more detailed modelling of the air flow dx.doi.org/10.1016/j.enconman.2010.03.016.
rate should also improve accuracy. Once this has been addressed, Peng, D., Robinson, D.B., 1976. A new two-constant equation of state. Ind. Eng.
Chem. Fundam. 15 (1), 59–64, http://dx.doi.org/10.1021/i160057a011.
refinements to the other model parameters can be made to reduce Quoilin, S., Aumann, R., Grill, A., Schuster, A., Lemort, V., Spliethoff, H., 2011.
the residuals further. Dynamic modeling and optimal control strategy of waste heat recovery
74 M.J. Proctor et al. / Geothermics 61 (2016) 63–74
organic rankine cycles. Appl. Energy 88 (6), 2183–2190, http://dx.doi.org/10. Vélez, F., Segovia, J.J., Martín, M.C., Antolín, G., Chejne, F., Quijano, A., 2012. A
1016/j.apenergy.2011.01.015. technical, economical and market review of organic rankine cycles for the
Schuster, A., Karellas, S., Kakaras, E., Spliethoff, H., 2009. Energetic and economic conversion of low-grade heat for power generation. Renew. Sustain. Energy
investigation of organic rankine cycle applications. Appl. Therm. Eng. 29 (8–9), Rev. 16 (6), 4175–4189, http://dx.doi.org/10.1016/j.rser.2012.03.022.
1809–1817, http://dx.doi.org/10.1016/j.applthermaleng.2008.08.016. Wang, T., Zhang, Y., Peng, Z., Shu, G., 2011. A review of researches on thermal
Sohel, M.I., Sellier, M., Brackney, L.J., Krumdieck, S., 2011. An iterative method for exhaust heat recovery with rankine cycle. Renew. Sustain. Energy Rev. 15 (6),
modelling the air-cooled organic rankine cycle geothermal power plant. Int. J. 2862–2871, http://dx.doi.org/10.1016/j.rser.2011.03.015.
Energy Res. 35 (5), 436–448, http://dx.doi.org/10.1002/Er.1706. Wei, D., Lu, X., Lu, Z., Gu, J., 2008. Dynamic modeling and simulation of an organic
Stepanoff, A.J., 1957. Centrifugal and Axial Flow Pumps: Theory, Design, and rankine cycle (ORC) system for waste heat recovery. Appl. Therm. Eng. 28 (10),
Application, 2nd ed. Wiley, New York. 1216–1224, http://dx.doi.org/10.1016/j.applthermaleng.2007.07.019.
Tchanche, B.F., Lambrinos, G., Frangoudakis, A., Papadakis, G., 2011. Low-grade Yari, M., 2010. Exergetic analysis of various types of geothermal power plants.
heat conversion into power using organic rankine cycles—a review of various Renew. Energy 35 (1), 112–121, http://dx.doi.org/10.1016/j.renene.2009.07.
applications. Renew. Sustain. Energy Rev. 15 (8), 3963–3979, http://dx.doi.org/ 023.
10.1016/j.rser.2011.07.024. Zhang, J., Zhang, W., Hou, G., Fang, F., 2012. Dynamic modeling and multivariable
Tempesti, D., Manfrida, G., Fiaschi, D., 2012. Thermodynamic analysis of two micro control of organic rankine cycles in waste heat utilizing processes. Comput.
CHP systems operating with geothermal and solar energy. Appl. Energy 97 (0), Math. Appl. 64 (5), 908–921, http://dx.doi.org/10.1016/j.camwa.2012.01.054.
609–617, http://dx.doi.org/10.1016/j.apenergy.2012.02.012.
Toffolo, A., Lazzaretto, A., Manente, G., Paci, M., 2012. An organic rankine cycle
off-design model for the search of the optimal control strategy. In: 25th
International Conference on Efficiency, Cost, Optimization, Simulation and
Environmental Impact of Energy Systems, Perugia, Italy, 295-1-295-14.
Vaja, I., Gambarotta, A., 2010. Internal combustion engine (ICE) bottoming with
organic rankine cycles (ORCs). Energy 35 (2), 1084–1093, http://dx.doi.org/10.
1016/j.energy.2009.06.001.