Professional Documents
Culture Documents
Renewable Energy: Luca Zanellato, Marco Astol Fi, Aldo Sera Fino, Dario Rizzi, Ennio Macchi
Renewable Energy: Luca Zanellato, Marco Astol Fi, Aldo Sera Fino, Dario Rizzi, Ennio Macchi
Renewable Energy
journal homepage: www.elsevier.com/locate/renene
a r t i c l e i n f o a b s t r a c t
Article history: The calculation of the performance of an ORC (Organic Rankine Cycle) and its components (in particularly
Received 29 January 2018 the turbine) is generally not a trivial task because of the lack of a proper instrumentation, the inaccuracy
Received in revised form of the available measurements and the uncertainty on fluid thermodynamic properties calculation. These
14 August 2018
limitations can greatly affect the final computed results especially for complex fluids with a small
Accepted 18 August 2018
Available online 20 August 2018
temperature drop in expansion. A strategy to solve the inconsistency of the measured data set is to verify
the energy and mass balance of each component and of the overall plant using also information related
to the geometry of specific components, primarily the turbine as well as electrical measurements. After a
Keywords:
Geothermal energy
brief description of the radial outflow turbine and of its main features, two operating geothermal ORC
Double level ORC (Organic Rankine Cycle) plants installed by Exergy Spa in Turkey are described. For both plants a
Experimental campaign methodology for the consistency check of the experimental data set is presented with a description of
Turbine efficiency the installed instrumentation, the test procedure and methods used to calculate the turbine efficiency as
well as the overall power cycle performance.
© 2018 Elsevier Ltd. All rights reserved.
https://doi.org/10.1016/j.renene.2018.08.068
0960-1481/© 2018 Elsevier Ltd. All rights reserved.
L. Zanellato et al. / Renewable Energy 147 (2020) 2896e2904 2897
Both the plants investigated in this paper have a double evap- 3. Experimental campaign and data set definition
oration levels but their plant layout substantially differ. Fig. 1 de-
picts their layouts and the instrumentation installed and used in 3.1. Description of the used instrumentation
the present study for the calculation of cycle performance and
turbine efficiency. As shown by Fig. 1, both plants are provided by an adequate field
The GREENECO plant exploits a 140 C geothermal brine and instrumentation for temperature and pressure measurements since
produces around 13 MW of electrical power. It is designed with two several standard precision probes are installed on both brine and
independent saturated cycles working with isopentane at two organic fluid streams. Thermodynamic state can be evaluated for
2898 L. Zanellato et al. / Renewable Energy 147 (2020) 2896e2904
Fig. 1. Plant schemes of GREENECO plant (left) and of the AKCA plant (right) with information about the field instrumentation. P and T indicate the point of measurements of
pressure and temperature respectively while W represents a power measure. Where symbol (*) is present an additional PT100 temperature instrument has been installed during the
experimental campaign because a field instrument was not installed.
Fig. 2. Schematic representation of the 4 stages high pressure turbine installed at GREENECO plant (left) and AKCA plant double inlet turbine (right). The two drawings are not
representative of the real dimensions of the two turbines.
Fig. 3. Aerial view of GREENECO (left) and AKCA (right) plants. Component PH2 for AKCA plant is not visible since it is placed below EVAHP.
L. Zanellato et al. / Renewable Energy 147 (2020) 2896e2904 2899
Table 1 Table 2
Redundant and missing field measurements in GREENECO and AKCA plants: (i) inlet Characteristics of the used instruments.
section, (o) outlet section.
Measured Standard Field Instrumentation
GREENECO quantity
Sensor Type Expanded
Redundant measurements uncertainty
screen. Both measures were collected for approximately 1 h set of 105 values are randomly generated with normal distri-
(including the 20 min sampling interval) and the overall energy is bution having a mean equal to the average value of the dataset
compared: the relative difference is 0.274% showing a high reli- and standard deviation equal to half of the instrument expanded
ability of the power output measured by the field instrument. uncertainty. The resulting values of turbine efficiency are nor-
It is important to note that verifying the energy balance for each mally distributed with an average value equal to 87.2% and an
component and evaluating their real performance, especially for expanded uncertainty (k ¼ 2) equal to ±8.2%. An even larger
the turbine, is a challenging task for at least four reasons: uncertainty is obtained for the other turbines characterized by a
lower temperature difference between inlet and outlet condi-
1. The Equation of State (EoS) implemented in Refprop 9.1 that tions like in AKCA plant and for low pressure GREENECO turbine.
links the measured pressure and temperature to the enthalpy 3. The poor frequency response of temperature instruments can
values, used as input for the energy balances and the efficiency lead to signal damping and time delay compared to pressure
calculation, is a complex function of 12 terms explicit in the instruments. To avoid this problem, stable operating conditions
reduced Helmholtz energy [9]. As result is not possible to are selected for the data sampling and time averaged value of
directly apply propagation of error rules to evaluate the uncer- each measured quantity is used for the further calculations.
tainty of a thermodynamic quantity starting from the measured Moreover, for some redundant measurements the average value
values of pressure and temperature. A possible strategy is to use between the two is used: this is the case of the temperatures
Montecarlo simulations for the evaluation of derived quantities between PH1HP/PH2HP and PHLP/EVALP in GREENECO plant
uncertainty. As reported in next point, this method results in 4. The possible inconsistencies between some measured quanti-
very large uncertainty on the results and it leads to a dramatic ties that cannot be detected when looking only to a specific
increase of computational time since some calculation requires component but they can be highlighted when considering the
an iterative procedure (i.e. calculation of choked mass flow rate). whole plant and the energy balance of all the components
2. The temperature instruments expanded uncertainty is almost in together.
the order of one Celsius degree; an uncertainty value that can be
adequate for steam cycles where temperature drop in the tur-
bine is few hundreds of Celsius degree while for ORC turbines it 4. Methodology
can involve a large uncertainty on the efficiency calculation. This
is mainly related to the use of a complex fluid and the presence In both GREENECO and AKCA power plants, the lack of some
of real gas effects at turbine inlet that reduces the expansion points of measurement and the absence of direct measures of mass
temperature drop and makes fluid thermodynamic properties flow rate do not allow closing the energy balance for each
and the turbine efficiency calculation very sensitive to both component of the cycle and performing a consistency check of the
outlet and inlet temperatures. A preliminary Montecarlo simu- experimental dataset. This is a typical issue in cycle performance
lation has been carried out for the efficiency calculation of the evaluation, especially for ORC where, due to the small plant size,
high temperature turbine of GREENECO plant starting from the installed instrumentation is strictly related to power plant
thermodynamic measured data. For the temperature and the control instead of performance monitoring. In addition, the inevi-
pressure measurements at turbine inlet and outlet conditions a tably presence of measures uncertainty may lead to relevant errors
in the evaluation of a single component performance if the
Fig. 4. Trend of data set collected in the experimental campaign. GREENECO is close to nominal point while for AKCA plant power production is lower than the nominal because of a
lower brine mass flow rate and temperature.
L. Zanellato et al. / Renewable Energy 147 (2020) 2896e2904 2901
calculation is not preceded by a data set consistency check through here is not suitable for performance tests to be performed in the
the verification of energy and mass balances. delivery phase of a ORC plant supply contract but, in authors
First, a check is performed between the experimental data and opinion, it is a valid strategy for the monitoring of plant compo-
the results of the state-of-the-art Equation of State. All the calcula- nents performance, to highlight malfunctioning and to schedule
tions are carried out assuming pure fluids and using Refprop 9.1 [9] maintenance related to heat exchangers cleaning or non-
for the calculation of fluid thermodynamic properties. The ther- condensable gases removal.
modynamic state of the fluid at evaporators outlet can be checked by
calculating the evaporation temperature from the value of measured 4.1. GREENECO plant
pressure and comparing it with the measured value of temperature.
In GREENECO plant it is found a small superheating at both high In order to calculate the thermodynamic properties of the
pressure and low pressure evaporators outlet, while for AKCA plant working fluid in each point of the plant, missing values of brine
no superheating is observed at both levels, in accordance with the pressure are evaluated assuming a homogeneous repartition of the
use of a demister. Same check is done at condenser hotwell overall pressure drop among the heat exchangers, allowing to
comparing the measured temperature with the saturation temper- calculate brine enthalpy. This assumption does not really affect the
ature at condenser inlet pressure (in case of the GREENECO high results since geothermal brine is modelled as a subcooled liquid
pressure plant we assumed a pressure drop on recuperator hot side with enthalpy almost independent of pressure. Same approach is
equal to 5%). Differently from GREENECO plant, in AKCA plant a small followed for the HP cycle where the total pressure drop between
subcooling is found: checked that the liquid level in the condensers pump outlet and evaporator inlet is assumed equally divided be-
was below the tube bundle we concluded that the “apparent” sub- tween REC, PH1HP and PH2HP. Organic fluid pump efficiency is set
cooling may be caused by the presence of non-condensable gas in equal to 0.75 considering manufacturer datasheet average perfor-
the condenser or marked pressure drops in the piping connecting mance and we verified that the effect of different assumed values is
the turbine and the condenser. However, this fact does not affect the negligible on the overall plant energy balance: varying the value of
methodology proposed and the calculus of derived quantities. pump efficiency by ±10% leads to variation lower than 0.15% on the
After these first calculations an iterative approach is defined for overall plant balance with a very small effect on final results. The
each plant in order to obtain the verification of the energy balance procedure used to calculate the plant performance is described and
of each component by slightly varying the values of some measured reported in Fig. 5.
quantities and by integrating all the information from the experi-
mental dataset with additional data related to turbine geometry. 1. HP fluid mass flow rate is estimated from the TURHP inlet con-
In particular, CFD calculations on turbine geometry can give ditions considering chocked flow in the outlet section of the 1st
useful information about the boundary layer blockage coefficient at stator vanes. Geometrical data of turbine (row radius, number of
blade throats and reference values for the calculated turbine effi- blades, gauging ratio, fillet factor) were measured by Exergy Spa
ciency and fluid velocities. In this work we refer to CFD calculations during turbine manufacturing and they are considered as input
performed by Exergy Spa with the commercial software Ansys CFX to the procedure. A coefficient equal to 0.95 is assumed to take
using a list of settings1 that, based on manufacturer experience, into account boundary layer blockage at blade throats. The
allows to estimate main quantities like mass flow rate and power calculated value of mass flow rate is in good agreement with the
output with an error lower than 1% respect to experimental data. result of an independent 3D CFD calculation made by Exergy on
Regarding the proposed methodology, it is important to un- the same turbine geometry.
derline four important aspects: (i) the consistency check is ob- 2. The same approach used at step 2 is followed for TURLP
tained by modification of the original data with variations that are component finding the low pressure mass flow rate coming
always smaller than the instrument expanded uncertainty, (ii) the from the low pressure evaporator EVALP
present approach does not allow to define the uncertainty of the 3. The overall energy balance of EVAHP and PH2HP components is
calculated results since only average data are used, however imposed since the measure of brine temperature between these
sensitivity analysis can be carried out to understand the impact of two components is not available. From the energy balance, and
different assumptions on the results, (iii) the methodology adopted neglecting heat losses, a first calculated value of the brine mass
depends on the available measures and the plant layout thus it flow rate is obtained. Successively, applying the energy balance
must be tuned on each specific case, (iv) the approach proposed for only EVAHP the brine temperature at evaporator pinch point
is found.
4. Energy balance of EVALP is imposed with the same assumptions
allowing to calculate a second value of brine mass flow rate.
1
Settings used in CFD calculations. 5. Energy balances of PH1HP and PHLP are imposed and a third
value of brine mass flow rate is calculated.
Steady state simulation and axisymmetric geometry.
6. Plant electrical power output is computed as the product of
Turbulence Model: SST (Shear Stress Transport)
Fluid model based on a Real Gas Property file created with NIST Refprop data mass flow rates and the corresponding enthalpy drops across
GGI (General Grid Interface) between rotor and stator rows is modelled with the turbines reduced by the mechanical and electrical generator
Stage Model (Mixing plane) losses. Shaft mechanical losses have been measured by Exergy
Convergence criteria: all RMS residual under 1e-06, power and mass flow rate
Spa in test bench and generator efficiency curve is provided by
variation below 1%.
Hexahedral mesh (Ansys Turbogrid mesh): Yþ is maintained in the range
generator manufacturer as function of power output.
30e300 in order to avoid the buffer layer. 7. As final step, three differences can be calculated: two between
Medium mesh is chosen with number of nodes z number of elements (2500 K) the calculated brine mass flow rates and the third between the
for each blade row. calculated and the measured electrical power output.
Mesh quality: face angle from 50 to 120 , edge length ratio below 50
Sensistivity analisys on three different meshes (1500, 2500 K and 3500 K ele-
ments): differences between medium and fine mesh on mass flow rate and In order to obtain the verification of the energy balance of each
turbine efficiency are below 1% component, the three differences are minimized by slightly varying
the thermodynamic values of the experimental dataset. It is found
that changing few brine and isopentane temperatures within a
2902 L. Zanellato et al. / Renewable Energy 147 (2020) 2896e2904
range of ±0.25 C respect the mean dataset value (a value lower 4.2. AKCA plant
than the instrument standard uncertainty) it is possible to reduce
all the three above mentioned differences down to less than 0.1%. For AKCA plant, it is not possible to calculate the turbine effi-
This allows verifying the energy balance of each component while ciency directly from the measures of temperature and pressure
taking in account the experimental data set. at inlet and outlet section since this machine has two different
The high pressure turbine mass flow rate computed with this inlets and the temperature between high and low pressure
procedure is 145.5 kg/s with a volume ratio through the expansion stages is not measured. Moreover, the two measured values of
equal to 6.54, while for the low pressure turbine these quantities turbine outlet temperature differ by around 2 C: a difference
are 129.2 kg/s and 3.03 respectively. Turbine total-to-static effi- that can dramatically affect the calculated turbine efficiency. As
ciencies are calculated with the following equation as function of a consequence, we assumed that both high pressure and low
only measured quantities ðTin ; pin ; Tout ; pout Þ. Turbine efficiency pressure stages have the same value of polytrophic efficiency as
result equal to 85.5% and 88.4% for high and low pressure machine representative of the whole machine average performance. A
respectively. guess value of turbine efficiency it is assumed at the beginning
of the process and it is varied during the procedure in order to
verify the plant overall energy balance. The following procedure
Dhreal hin ðTin ; pin Þ hout ðTout ; pout Þ is used and the schematic flowchart is shown in Fig. 6:
hturb ¼ ¼
Dhideal hin ðTin ; pin Þ hisout ðsin ðTin ; pin Þ; pout Þ 1. The high pressure mass flow rate is calculated as previously
explained for GREENECO plant (step 2) knowing the inlet ther-
modynamic conditions and the geometrical data of the first
stator row.
Considering that at turbine discharge the flow in the diffuser 2. Thanks to the use of an additional high precision temperature
may be not homogenously distributed a sensitivity analysis is car- probe installed at the outlet section of the high pressure evap-
ried out for GREENECO plant with the aim to understand the impact orator, the imposition of the energy balance of EVAHP compo-
of the turbine outlet temperature on the verification of the energy nent allows to calculate the geothermal brine mass flow rate
balances and the calculated turbine efficiencies. The procedure is that is used for all the following calculations.
repeated assuming the turbine discharge temperature equal alter- 3. The energy balance of PH2 is checked using the organic fluid
natively to the lower and to the higher measured value (respec- mass flow rate calculated at step 1 and the brine mass flow rate
tively at turbine discharge and condenser inlet) while for the other calculated at step 2. The relative difference on the energy bal-
data a maximum variation equal to the measurement expanded ance is calculated.
uncertainty is considered. The three minimized differences are 4. The evaluation of the mass flow rate coming from the low
higher than in the previous case but always below 1.5%. The pressure evaporator is less straight forward than for the
maximum efficiency of TURHP and TURLP are respectively equal to GREENECO case and a slightly more complex procedure is
86.4% and 88.65% while the minimum ones are 84.3% and 88.1%. required:
i. First limit is represented by the lack of the thermowell for
the temperature measurement at EVALP outlet that does not
allow to directly calculating the organic fluid mass flow rate
from the component energy balance. A guess value of the
pinch point temperature difference (DTpp ) at EVALP is
assumed. With this information and by assuming the brine
mass flow rate equal to the value calculated at step 2 a first
value for the mass flow rate coming from the low pressure
evaporator is calculated.
ii. Second limit is due to the circular reference that links the
chocked mass flow rate at second stator throat to the ther-
modynamic state at the inlet of the second stator row which
depends itself from the mass flow rate coming from EVALP.
To solve this issue the thermodynamic state at the first stage
outlet is firstly calculated using the guess value of turbine
efficiency. The second stage inlet thermodynamic conditions
are hence calculated assuming a perfect mixing between the
low pressure vapor from EVALP calculated at step 4.i and the
fluid expanded from high pressure turbine first stage
calculated at step 1. The total mass flow rate expanded by
low pressure stages is calculated knowing the second stator
row geometrical data. A second value of mass flow rate
coming from the low pressure evaporator is finally calcu-
lated subtracting from the total mass flow rate the mass flow
rate coming from high pressure turbine.
iii. The difference between the low pressure mass flow rates
calculated at step 4.i and 4.ii is calculated and it is set to zero
by varying the assumption on DTpp .
5. The mass and energy balance of brine mixing process after
EVALP is imposed allowing to calculate the mass flow rate of the
brine coming from the steam condenser and the mass flow rate
Fig. 5. Flow chart of the methodology adopted for GREENECO plant. of brine entering in the PH1 component.
L. Zanellato et al. / Renewable Energy 147 (2020) 2896e2904 2903
Fig. 7. T-Q diagrams for the heat introduction process for GREENECO and AKCA plants. The red dot in Akca plant represent the temperature of the brine stream coming from the
steam condenser. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)
2904 L. Zanellato et al. / Renewable Energy 147 (2020) 2896e2904
Table 3 to calculate the missing measured values and to obtain a values for
Main quantities calculated for GREENECO and AKCA plant. the calculated quantities that allow to respect the energy balance of
GREENECO AKCA each component. The procedure starts from average data of the
Turbine efficiency HP 84.3%e85.5% - 86.4% 88.8%e91.9% -96.1%
experimental campaign and, by applying small modification to the
LP 88.1%e88.4% - 88.65% initial data, reaches a consistent solution. This approach relies on a
heat input from geothermal HP 61.5 28.2 number of assumptions that can affect the final results and requires
brine, MW LP 54.4 the availability of components manufacturer in sharing sensible
Turbine mechanical power, HP 8.38 0.5
data like blade geometry and equipment specification. The uncer-
MW LP 4.57 2.41
Electric power output, MW 12.68 2.83 tainty associated to the derived value is not computed and the
Plant efficiency 10.8% 10.0% calculated values should be considered as a first, but reasonable,
Second law efficiency* 58.1% 68.2% estimation of real quantities. Additional sensitivity analysis on both
evaporator DTpp , C HP 1.18 1.35
the assumptions and the reliability of measurements can be carried
LP 1.11 1.60
out in order to understand their impact on the calculated values.
The procedure has been successfully applied to two plants differing
in both the plant configurations and the number and location of
to adopt large heat exchanger surface in order to increase system measurement points allowing to calculate values of isentropic ef-
efficiency. In example, GREENECO plant efficiency results equal to ficiency that are warranted on the basis of a large set of experi-
10.8% that corresponds to a 58.1% second law efficiency calculated mental data, including direct electrical power measurement, and
respect to the trapezoidal Lorenz cycle working between brine information on turbine row geometry.
temperature and condensation temperature, while for AKCA the
plant and the second law efficiencies are 10.0% and 68.2% respec-
tively. Power output of GREENECO plant is very close to the design
References
value while for AKCA plant the marked penalization of power
output is due to the provisional presence of non-condensable gases [1] E. Macchi, M. Astolfi, Organic Rankine Cycle (ORC) Power Systems, Elsevier,
in the condenser. Table 3 reports the main quantities calculated for 2017. Woodhead Publishing Series in Energy Number 107.
the two plants using time averaged values of the measured quan- [2] E. Macchi, M. Astolfi, The choice of working fluid: the most important step for a
successful organic rankine cycle (and an efficient turbine), in: Proc. Of 2nd Int.
tities while for the turbine efficiency both minimum and maximum Sem. on ORC Power Systems, Oct., Rotterdam, 2013.
values obtained from the sensitivity analysis are shown. [3] C. Spadacini, D. Rizzi, C. Saccilotto, S. Salgarollo, L. Centemeri, The radial outflow
turbine technology: impact on the cycle thermodynamics and machinery fluid-
and rotordynamic features, in: Proc. Of 2nd Int. Sem. on ORC Power Systems,
6. Conclusions Oct., Rotterdam, 2013.
[4] C. Spadacini, L. Centemeri, L.G. Xodo, M. Astolfi, M.C. Romano, E. Macchi, A new
In principle, the turbine isentropic efficiency could be simply configuration for organic rankine cycle power systems, in: Proc. Of 1st Int. Sem.
on ORC Power Systems, Sept., Delft, 2011.
computed from values of inlet and outlet thermodynamic condi- [5] M. Pini, G. Persico, E. Casati, V. Dossena, Preliminary design of a centrifugal
tions of the working fluid. However, this procedure cannot be turbine for ORC applications, ASME J. Eng. Gas Turb. Power Apr. 115 (2013).
applied for steam turbines, since the expansion usually ends inside [6] G. Persico, M. Pini, V. Dossena, P. Gaetani, Aerodynamic design and analysis of
centrifugal turbine cascade, in: ASME Turbo Expo 2013, 2013. Paper No.
the saturation line and is not possible to measure the steam quality GT2013-95770.
at outlet conditions. Also in open cycle gas turbines the procedure is [7] C. Spadacini, L. Centemeri, D. Rizzi, M. Sanvito, A. Serafino, Fluid-dynamics of
not possible, since inlet turbine temperature are too high for the ORC radial outflow turbine, in: Proc. Of 3rd Int. Sem. on ORC Power Sys-
tems, Oct., Brussels, 2015.
temperature probes and the expansion is far from adiabatic, being
[8] M. Frassinetti, D. Rizzi, A. Serafino, L. Centemeri, C. Spadacini, Operational re-
strongly cooled. In ORC turbines the procedure is possible, but the sults of the World's first orc radial outflow turbine, and its future development,
small temperature drop across the turbine can leave doubts about in: Proc. Of 2nd Int. Sem. on ORC Power Systems, Oct., Rotterdam, 2013.
the precision of the calculated value. This paper proposes a meth- [9] E.W. Lemmon, M.L. Huber, M.O. McLinden, NIST Standard Reference Database
23: Reference Fluid Thermodynamic and Transport Properties-refprop, Version
odology to solve this issue by the analysis of the whole cycle in 9.1, National Institute of Standards and Technology, Standard Reference Data
order to perform a consistency check on the experimental dataset, Program, Gaithersburg, 2013.