You are on page 1of 10

THE IMPELLER EXIT FLOW COEFFICIENT AS A PERFORMANCE MAP VARIABLE FOR

PREDICTING CENTRIFUGAL COMPRESSOR OFF-DESIGN OPERATION APPLIED TO A


SUPERCRITICAL CO2 WORKING FLUID

Eric Liese Stephen E. Zitney


National Energy Technology Laboratory, National Energy Technology Laboratory,
Department of Energy – Morgantown, WV, USA Department of Energy – Morgantown, WV, USA
eric.liese@netl.doe.gov stephen.zitney@netl.doe.gov

ABSTRACT 𝑀𝑈2 Impeller tip Mach number (using inlet speed of sound)
A multi-stage centrifugal compressor model is presented nT Polytropic temperature exponent
with emphasis on analyzing use of an exit flow coefficient vs. an m Polytropic temperature exponent ratio
inlet flow coefficient performance parameter to predict off- R Universal gas constant
design conditions in the critical region of a supercritical carbon ∆ℎ𝑟𝑜𝑡𝑜𝑟
Rxn Stage reaction: 𝑅 =
dioxide (CO2) power cycle. A description of the performance ∆ℎ𝑠𝑡𝑎𝑔𝑒

parameters is given along with their implementation in a design 𝑈2 Impeller tip velocity
model (number of stages, basic sizing, etc.) and a dynamic model 𝑉3̇ ⁄𝑉𝑡𝑖𝑛
̇ Impeller exit static volume/Inlet total volume flow rate
(for use in transient studies). A design case is shown for two Z Compressibility factor
compressors, a bypass compressor and a main compressor, as
defined in a process simulation of a 10 megawatt (MW) Greek
supercritical CO2 recompression Brayton cycle. Simulation 𝜇𝑝 Polytropic head coefficient: 𝜇𝑝 = 𝜂𝑝 ∗ 𝐼
results are presented for a simple open cycle and closed cycle 𝜂𝑝 Polytropic efficiency
process with changes to the inlet temperature of the main 𝜂𝑝3 Impeller total-to-static polytropic efficiency
compressor which operates near the CO2 critical point. Results 𝜂𝑠 Isentropic efficiency
showed some difference in results using the exit vs. inlet flow 𝜂𝐿 Leakage efficiency
coefficient correction, however, it was not significant for the ρ Density
range of conditions examined. This paper also serves as a 𝑉̇𝑖𝑛
𝜑 Flow coefficient: 𝜑 = 𝜋
reference for future works, including a full process simulation of 𝑑 2𝑈
4 2 2
the 10 MW recompression Brayton cycle. 𝑉3̇
𝜑3 Impeller exit flow coefficient: 𝜑3 =
𝜋𝑑2 𝑏2 𝑈2
1. NOMENCLATURE
ain Speed of sound at compressor inlet Subscripts
𝑏2 Impeller exit width 2 Impeller exit station
𝑑2 Impeller exit diameter 3 Impeller exit static condition
𝑓 Recirculation leakage factor s Isentropic designation
𝐹̇ Molar flow rate tin Total inlet condition
𝐹𝑓̇ Recirculation molar flow rate
∆ℎ
I Impeller work input factor: 𝐼 =
𝑈2 2
2. INTRODUCTION
I *
Impeller slip factor There is significant renewed interest in power cycles that
𝜈 𝜕𝑃 use supercritical CO2 (sCO2) as a working fluid. The two
𝑘𝜈 Isentropic volume exponent: 𝑘𝜈 = − ( ) primary cycle types are “indirect” - where a heat exchanger is
𝑃 𝜕𝜈 𝑠
𝑘𝑇 Isentropic temperature exponent used to transfer heat between a heating gas and the CO2 which is
𝑚̇ Mass flow rate in a closed loop, and “direct” - where a fuel and oxygen are

1
combusted to produce the CO2 working fluid and water. An Maps are given in both dimensional and dimensionless
initial collaborative effort between the National Energy form. For dimensional maps, the x-axis is always a flow variable
Technology Laboratory and the University of Genoa used (mass or volume flow) and the y-axis is typically either the
Genoa’s MATLAB-based dynamic modeling environment to pressure ratio or the “head”, that is, the enthalpy rise (actual,
model a 10MW indirect s CO2 recompression Brayton cycle [1]. isentropic, or polytropic enthalpy rise). The efficiency is also
This paper showed interesting results with respect to the plotted against the flow rate, either along with the pressure ratio
significant effect a slight temperature change of several degrees or head on the same map or separately. Sometimes maps are also
Kelvin at the main compressor inlet can have on the overall presented in dimensionless parameters, such as the head
system temperature profile. The compressor model in [1] uses coefficient (or work input factor) and flow coefficient. This is
corrected performance maps for a real gas, but it assumes a more common with centrifugal compressors and this idea will
constant specific heat ratio, which is common; however, this become clearer since this work uses dimensionless parameters.
assumption is not necessarily valid in the critical region where While the map and standard thermodynamic relations can
fluid properties change dramatically. Therefore, it is desirable be used to determine the compressor work output and exit fluid
to have a compressor model with performance parameters that state of the compressor, it is only valid for a given inlet fluid
account for a variable specific heat ratio, or better described, a state. However, under certain assumptions it can be shown that
variable volume exponent. a map can be used (or generated) with appropriate inlet state
The centrifugal compressor model developed in this study is corrections. For an axial compressor using an ideal gas, these
described in two parts: 1) the design model in Aspen Custom corrections are derived from a comparison of the Euler
Modeler (ACM) for use in determining the basic compressor turbomachinery equation and the conservation of mass equations
dimensions (including the number of stages) and 2) the off- [3]. This approach can also be extended to a real gas, and these
design model for use in building the sCO2 cycle model in Aspen corrections are discussed and shown in [1]. However, it is
Plus Dynamics. Both are based on an ACM model originally important to point out that the corrections used in [1] assume a
developed by Modekurti et al. [2] for use in post-combustion constant specific heat ratio, the ratio of the heat capacity at
CO2 capture and storage applications. The modifications of the constant pressure to heat capacity at constant volume. For a
multi-stage compressor in [2] for use in s CO2 power cycles will centrifugal compressor, the Buckingham Pi theorem is typically
be detailed in this paper. A design is performed for the main used to compute the dimensionless parameters from the given
compressor and bypass compressor given the steady-state design variables. A couple of good papers that describe the parameters
conditions for the 10 MW recompression Brayton cycle shown that come from the dimensional analysis for centrifugal
in [1]. Performance results are given at the design point for all compressors are by Boyce [4] and Brown [5]. Another artifact
the pertinent variables. Results are also shown for off-design that falls out of the dimensional analysis is the fan law as
conditions and the results of operating the main compressor near described by Brown [5]. Instead of using separate speed lines
the critical point are highlighted. First, some discussion of (and thus needing to interpolate between them in a computational
compressor performance map parameters is given to familiarize environment), a single line can be used and is adjusted for
readers with their origin and use, as well as to set up the varying speed. Brown [5] demonstrates this as a fairly accurate
motivation for developing the custom multi-stage compressor representation given the simplification. Other researchers also
model. show similar results [e.g., 6]. Again, a constant specific heat
ratio is assumed. Interestingly, the performance map corrections
3. PERFORMANCE MAP PARAMETERS AND used for both axial and centrifugal compressors are the same
CORRECTING FOR VARIABLE INLET CONDITIONS even though the analysis is slightly different.
The compressor model uses performance maps to determine In the thesis by Baltadjiev [7], it is shown that the inlet state
the relationship between turbomachinery work, speed, and fluid corrections discussed above could become more incorrect as the
states. For instance, if the flow rate and compressor speed are CO2 critical point is approached, especially at an increasingly
known, the compressor efficiency, work, exit pressure, etc. can higher Mach number. This degradation in the accuracy of the
be determined. As is well known, there are numerous ways in predictions occurs in part since the derivation of the corrections
which performance maps can be posed and generated. The most as shown in [7] is based on a constant inlet isentropic volume
desirable maps are ones generated from testing of the exact exponent, kν (ns in Baltadjiev). This is analogous to the
compressor that is to be used in the system. If unavailable, then discussion above regarding the use of constant specific heat ratio
a map of a similar compressor type can be scaled to the desired for compressor performance corrections. Thus, Baltadjiev
design conditions. In this work, as will be described later, a suggests an empirical safety boundary to where it is acceptable
generalized map is derived from basic compressor sizing along to assume the standard corrections (the region outside the safe
with some input from a given map that is scaled. zone is called here the “critical region”). In a sCO2
recompression Brayton cycle, the turbine expander and bypass

2
compressor inlet state conditions are in this safe zone (although In the following, we will discuss differences between the current
perhaps not at startup); however, the main compressor is not, model and the model developed in [2].
with inlet conditions near to the critical point. Thus, using
typical corrections for the main compressor is more circumspect, 4. DESIGN MODEL
especially if the inlet kν varies greatly from its design point. To the authors' knowledge, compressors for a 10 MW
A typical dimensionless map will use parameters which recompression Brayton system do not yet exist, nor do their
have the corrections for inlet conditions built in, for example proposed designs (i.e., sizes and maps). Thus, a simple design
plotting the work input factor (or head coefficient), 𝐼, vs. the flow model is needed and described here. For a given fluid with a
coefficient, 𝜑, especially for centrifugal turbomachinery. These given inlet temperature, pressure, and flow rate condition and
parameters are defined as: desired outlet pressure, a design point for the map(s) is
determined, that is, the exit flow coefficient, 𝜑3 , the slip
∆ℎ coefficient, 𝐼 ∗ , and polytropic efficiency, 𝜂𝑝 . The design model
𝐼= (1)
𝑈2 2 is also used to set up initial conditions for the leakage and friction
losses, which will be described later. Notice that to calculate the
̇
𝑉𝑡𝑖𝑛
𝜑=𝜋 2 (2) flow coefficients described above, the impeller wheel (outer)
𝑑 𝑈 diameter, 𝑑2 , and the impeller exit width (or impeller blade tip
4 2 2
height), 𝑏2 , are calculated. The number of stages is determined
where Δh is the static enthalpy rise across the compressor, 𝑑2 is by design constraints and a requirement to operate the
the impeller diameter, and 𝑈2 is the impeller tip speed, where the compressor near maximum efficiency. This multi-stage
subscript 2 denotes the impeller exit station (given by Ludtke compressor model is described in more detail by Modekurti et
[8]). al. [2]. Note that [2] determines an expected design point
Ludtke [8] uses the variable, s, for the work input factor, efficiency based on Figures in Aungier [9] (Chapter 6). In the
while 𝐼 is used here to be consistent with [2] and [9]. (Note that extended model here, a calculation of 𝑏2 is added (it is an
sometimes the Greek letter 𝛹 is used and called the head assumed value by [2]). It is based on an average of a simplified
coefficient, and Ludtke does use 𝛹𝑝 to denote the polytropic head empirical relationship between 𝜑3 and 𝜑 that is suggested by
coefficient. Aungier [9] uses 𝜇𝑝 for the polytropic head Ludtke [8] (Section 6.2.2, Fig. 6.10) to obtain reasonable
coefficient. Also, Ludtke defines 𝛹𝑝 = 2 ∗ 𝜂𝑝 ∗ 𝐼 whereas impeller dimensions. It is given here as:
Aungier defines 𝜇𝑝 = 𝜂𝑝 ∗ 𝐼, thus making the difference a 𝜑3 = 𝜑 + 0.215 𝑓𝑜𝑟 𝜑 < 0.038 (4)
factor of two. The point is that the nomenclature can 𝜑3 = 1.8𝜑 + 0.187 𝑓𝑜𝑟 𝜑 ≥ 0.038 (5)
unfortunately be easily confused and misread. Here, we use the
Aungier definition since it is also used in [2]). The relations above are used along with the equation:
The flow coefficient is a non-dimensionalized inlet volume
flow. However, this assumes a constant inlet temperature and 𝑏2 𝜑(𝑉3̇ ⁄𝑉𝑡𝑖𝑛
̇ )𝑓
= (6)
specific heat ratio as noted by Ludtke [8] (Section 5.4). Ludtke 𝑑2 4𝜑3 𝜂𝐿
demonstrates that if a proposed exit flow coefficient is used
instead of the inlet flow coefficient, then changes in inlet where 𝑓, 𝜂𝐿 and 𝑉3̇ ⁄𝑉𝑖𝑛̇ will be defined and calculated in the
temperature, real gas effects, and changes to the isentropic following paragraphs.
volume coefficient can also be accounted for in the performance The design model setup also calculates a design value for
map. The exit flow coefficient is defined as: 𝜑3 along with 𝐼 ∗ . The reason for using the slip coefficient, 𝐼 ∗ ,
instead of the work input factor, 𝐼, will be explained in the
𝑉3̇ (3)
𝜑3 = discussion of the off-design model in Section 5. The relationship
𝜋𝑑2 𝑏2 𝑈2 between 𝐼 and 𝐼 ∗ pertains to the losses. The relationship is given
by Ludtke [8] as:
where the subscript 3 denotes the static condition at an
infinitesimal distance downstream of the impeller exit, and thus 𝐼∗ 𝐶𝑀 𝜂𝐿 1
𝑉3̇ is the static volumetric flow rate at the impeller exit. = 𝜂𝐿 − (7)
𝐼 2𝜋 𝑓 𝐼 ∗ 𝜑(𝑉3̇ ⁄𝑉𝑡𝑖𝑛
̇ )
The use of the Ludtke approach [8] is described, in part, by
Modekurti et al. [2] and is used in [2] to better model a
where 𝜂𝐿 is the shroud leakage efficiency, 𝑓 is the recirculation
compression train used for CO2 sequestration. Modekurti et al.
also uses the Aungier [8] approach for much of the design model. leakage factor, 𝐶𝑀 is the disk friction loss coefficient, and 𝑉3̇ ⁄𝑉𝑖𝑛
̇
is the ratio of the impeller exit static volumetric flow rate to the
inlet total volumetric flow rate.

3
The values for the shroud and recirculation losses, which 𝑛 𝑇3 − 1 𝑍𝑡𝑖𝑛 𝑅 1 𝑘 𝑇3 − 1
will be seen in the results, are taken from correlation estimates 𝑚= = ( − 1) + (13)
𝑛 𝑇3 𝑐𝑝 𝑖𝑛 𝜂𝑝3 𝑘 𝑇3
given by Ludtke [8]. The recirculation leakage factor is
calculated for the last stage only and is based on: 𝑇3𝑠
𝑘 𝑇3 − 1 𝑙𝑛 (𝑇𝑡𝑖𝑛 )
𝑚̇ + ∆𝑚̇𝑟𝑒𝑐𝑖𝑟𝑐 = (14)
𝑘 𝑇3 𝑃
𝑓= (8) 𝑙𝑛 ( 3 )
𝑚̇ 𝑃𝑡𝑖𝑛

where ∆𝑚̇𝑟𝑒𝑐𝑖𝑟𝑐 is the added recirculation leakage flow √𝜂𝑝 + 𝑅𝑥𝑛 − 1


𝜂𝑝3 = (15)
𝑅𝑥𝑛
The shroud leakage efficiency is calculated using:
𝜑3 − 𝐼 2
𝑅𝑥𝑛 = 1 − (16)
𝑓𝑚̇ 2∗𝐼
𝜂𝐿 = (9)
𝑓𝑚̇ + ∆𝑚̇𝑠ℎ𝑟𝑜𝑢𝑑
and where kT3 is the isentropic temperature coefficient, Rxn is
where ∆𝑚̇𝑠ℎ𝑟𝑜𝑢𝑑 is the shroud leakage flow. the stage reaction, and 𝜂𝑝3 is the impeller total-to-static
In addition, for off-design conditions, the ∆𝑚̇ in each efficiency approximation by Ludtke [8].
equation is adjusted based on the relationship: It remains to give a relationship for either the pressure ratio
P3/Ptin. Ludtke [8] derives this for both an ideal and real gas.
∆𝑚̇𝑜𝑓𝑓 𝑑𝑒𝑠𝑖𝑔𝑛 √∆𝑃𝑜𝑓𝑓 𝑑𝑒𝑠𝑖𝑔𝑛 First, the polytropic total-to static head (enthalpy change) from
= (10) the Euler turbomachinery equations resolves to be:
∆𝑚̇𝑑𝑒𝑠𝑖𝑔𝑛 √∆𝑃𝑑𝑒𝑠𝑖𝑔𝑛
𝜂𝑝3 ∗ 𝑅𝑥𝑛 ∗ 𝐼 ∗ ∗ 𝑈2 (17)
where the ∆𝑃 is between the impeller exit and inlet for the shroud
loss, and between the last stage impeller exit and first stage inlet
Next the polytropic head is also defined by:
for the recirculation loss. 𝑛𝑣3 −1
The disc friction coefficient, 𝐶𝑀 , is assumed constant for all 𝑛𝑣3 𝑃3 𝑛𝑣3
𝑍𝑡𝑖𝑛 𝑅 ∗ 𝑇𝑡𝑖𝑛 [ − 1] (18)
conditions. It does have a Reynolds number relationship that is 𝑛𝑣3 − 1 𝑃𝑡𝑖𝑛
described by both Ludtke [8] and Aungier [9]; however, after
trying these it was found that a constant value was sufficient for
where the impeller polytropic volume coefficient is:
most of the operating range given the compressor designs here.
𝑘𝑣3
Finally, a calculation of the ratio 𝑉3̇ ⁄𝑉𝑡𝑖𝑛
̇ is needed for Eqns. 𝑛𝑣3 =
1 𝑘 𝑇3 − 1
6 and 7, as well as for the exit flow coefficient, 𝜑3 , in Eqn. 3. 1 + (1 − ) 𝑘𝑣3 (19)
𝜂𝑝3 𝑘 𝑇3
The calculation of this ratio here differs from the model in [2]
and follows more closely the equations given by Ludtke [8] and the impeller isentropic volume coefficient is:
(Section 6.2.2). The general equation used here is:
𝑃3
𝑉3̇ 𝑇3 𝑃𝑡𝑖𝑛 𝑍3 𝑙𝑛 ( )
𝑃𝑡𝑖𝑛
= (11) 𝑘𝑣3 = 𝜌 (20)
̇
𝑉𝑡𝑖𝑛 𝑇𝑡𝑖𝑛 𝑃3 𝑍𝑡𝑖𝑛 𝑙𝑛 ( 3𝑠 )
𝜌𝑡𝑖𝑛
where 𝑉𝑡𝑖𝑛̇ is the total volumetric flowrate at the inlet, 𝑍𝑡𝑖𝑛 and
Equations (17) and (18) are equated to solve for the pressure
𝑍3 are compressibility factors at the inlet and impeller exit,
ratio. Once the exit impeller static pressure and temperature are
respectively, 𝑃𝑡𝑖𝑛 and 𝑃3 are pressures at the inlet and impeller
calculated, the compressibility factor, Z, may be obtained using
exit, respectively, and 𝑇𝑡𝑖𝑛 and 𝑇3 are temperatures at the inlet a physical properties routine. The REFPROP method by the
and impeller exit, respectively. National Institute of Standards and Technology is used here [10].
The temperature and pressure relation for a real gas is given This gives all the relationships needed to make all the
by the following: necessary calculations. It is apparent that the problem is iterative
since the reaction contains 𝜑3 and furthermore the isentropic and
𝑇3 𝑃3 𝑚
= ( ) (12) polytropic coefficients depend on knowing the inlet and exit
𝑇𝑡𝑖𝑛 𝑃𝑡𝑖𝑛 impeller conditions.
where

4
5. OFF-DESIGN MODEL – COMPRESSOR MODEL Surge Point
USED IN SYSTEM SIMULATION
Currently, there is also no rotational inertia or shaft speed
dynamic modeled, which is straightforward to compute from a
physics point of view. However, the inertia term can be difficult
to know unless the actual equipment (e.g., the motor) and setup
(e.g., separate or common shaft for the turbomachinery) is
specified, and furthermore it is not uncommon to ignore the shaft
dynamics when looking at many process dynamic scenarios.
With the previous design setup parameters given to the off-
design compressor model, the next steps are to define the
compressor characteristic map in terms of the performance
parameters, and to calculate the corrected exit flow, 𝜑3 , based
on the inlet flow conditions (from the equations given in Section
4 on the Design Model). Then the efficiency, 𝜂𝑝 , and slip Figure 1. Compressor performance map
coefficient, 𝐼 ∗ , can be determined from the map. Ultimately the
work coefficient, compressor work and outlet fluid states can be
calculated using standard thermodynamic relations. As in the 6. SURGE MARGIN ASSUMPTION
design problem, the solution of the off-design model is an Knowing under what conditions surge occurs during
iterative process. compressor operation (and staying away from the event) is
Our main reference, Modekurti et al. [2], uses scaled vendor important. For a given compressor that has been experimentally
maps. Here, the approach described in Ludtke [8] (Section 6.8) mapped, the surge point is known, and it is usually the minimum
is used where 𝜑3 and 𝐼 ∗ are defined to have a linear relationship flow coefficient plotted, or if used, the minimum corrected flow
and are related to the design points as shown: rate. That is, it is the left most point on the map. Since this study
uses the exit flow coefficient, it brings up an interesting question
𝜑3 ∗ as to whether the surge condition should also be based on this
𝐼∗ = 1 − (1 − 𝐼𝑑𝑒𝑠𝑖𝑔𝑛 ) (21)
𝜑3 𝑑𝑒𝑠𝑖𝑔𝑛 parameter, or if it should still be based on the inlet flow
coefficient condition. Both surge conditions will be monitored
Then, the equations described in the Design Model discussion and compared in the results.
(Section 4.0 above) are used to determine 𝐼. So far there has been no determination of the exact location
An efficiency is needed along with the energy equation and of that point for the model(s) here. A general approach is not
isentropic relations to determine the exit temperature and possible since this depends on many specifics of the compressor
pressure. The efficiency map used here is scaled from a map as design. The best the authors can do at present is to scale the map
described in [2]. The same generalized equation is used for both of the last and highest pressure stage CO2 compressor shown by
the main and bypass compressor and is given as: Modekurti [2]. Using this map, the surge point with respect to
the design point is:
2
𝜑3
𝜂𝑝 = 𝜂𝑝 𝑑𝑒𝑠𝑖𝑔𝑛 ∗ [− 0.828 ( ) 𝜑3 𝑠𝑢𝑟𝑔𝑒 = 0.7216𝜑3 𝑑𝑒𝑠𝑖𝑔𝑛 (23)
𝜑3 𝑑𝑒𝑠𝑖𝑔𝑛
(22)
𝜑3 where 𝜑3 is replaced by 𝜑 for the surge point based on the inlet
+ 1.453 + 0.374] flow coefficient. Note that this equation is used for each stage
𝜑3 𝑑𝑒𝑠𝑖𝑔𝑛
in the multi-stage compressor models presented in this study.
A plot of Eqns. (21) and (22) is shown in Fig. 1 as the While the exact point of surge is unknown. The surge margin is
relative exit flow coefficient vs. the relative polytropic defined as a percentage by:
efficiency, ηpr, and relative impeller slip factor, I*r, where
relative means divided by the design condition and assuming a 𝜑3 𝑠𝑢𝑟𝑔𝑒 − 𝜑3
𝑆𝑢𝑟𝑔𝑒 𝑀𝑎𝑟𝑔𝑖𝑛 = ∗ 100 (24)

case for 𝐼𝑑𝑒𝑠𝑖𝑔𝑛 = 0.69. This is the fundamental compressor 𝜑3
performance map. Note that the design point, 1, is slightly to the
where 𝜑3 is replaced by 𝜑 for the surge margin based on the inlet
right of the efficiency peak and the efficiency decrease as the
flow coefficient.
surge point is approached is modest (optimistic) as given here.
The assumed surge point is discussed in the following section.

5
7. RESULTS nv3 5.29 6.41
The multi-stage compressor design model described above m3 0.092 0.102
was implemented and simulated in Aspen Custom Modeler V3/Vin 0.918 0.958
(ACM). Tables 1, 2 and 3 list many of the relevant design P3 [bar] 129.7 208.0
parameters for the main compressor and Tables 4, 5, and 6 for T3 [K] 320.4 336.2
the bypass compressor. The main compressor design ended up ρ3 [kg/m3] 667.4 717.1
having two stages; however, a three-stage design could have also
been chosen, which would have given a slightly higher Table 4. Bypass compressor stage inlet and outlet fluid
efficiency. The two-stage design was used since the efficiency states. 𝑚̇ = 35.6 𝑘𝑔/𝑠 for all stages
predicted by the design model matched well with the current Stage 1 Stage 2 Stage 3
overall efficiency used in the system design. As mentioned, the Pin [bar] 85.4 126.6 178.6
design model here utilizes the Modekurti et al. model [2] with Pout [bar] 126.6 178.6 240.79
modifications. The same impeller diameter is used for all stages. Tin [K] 351.8 389.1 424.0
Tout [K] 389.1 424.0 456.3
Table 1. Main compressor stage inlet and outlet fluid states ρin [kg/m3] 177.4 231.3 285.8
𝑚̇ = 64.4 𝑘𝑔/𝑠 for all stages ρout [kg/m3] 231.3 285.8 337.6
Stage 1 Stage 2 Zin 0.7238 0.7448 0.7800
Pin [bar] 85.0 158.9 Δh [kJ/kg] 24.79 25.38 25.97
Pout [bar] 159.0 239.2
Tin [K] 308.15 327.20 Table 5. Design parameter results for bypass compressor
Tout [K] 327.20 341.68 𝑑2 = 0.1951 m
ρin [kg/m3] 612.87 686.56 Stage 1 Stage 2 Stage 3
ρout [kg/m3] 686.56 730.69 𝜂𝑠 0.807 0.784 0.762
Zin 0.2384 0.3745 𝜂𝑝 0.816 0.792 0.770
Δh [kJ/kg] 13.56 13.68 I* 0.682 0.689 0.698
I 0.734 0.751 0.768
Table 2. Design parameter results for main compressor. 𝜑 0.0365 0.0280 0.0227
𝑑2 = 0.1455 m 𝜂𝐿 0.963 0.959 0.956
Stage 1 Stage 2 𝐹𝜂̇
𝐿 𝑑𝑒𝑠𝑖𝑔𝑛
[kmol/hr] 111 125 133
𝜂𝑠 0.828 0.822 ∆𝑃𝜂𝐿 𝑑𝑒𝑠𝑖𝑔𝑛 [bar] 24.2 31.1 37.6
𝜂𝑝 0.833 0.825 𝑓 - - 1.035
*
I 0.694 0.692 𝐹𝑓̇ 𝑑𝑒𝑠𝑖𝑔𝑛 [kmol/hr] - - 102
I 0.722 0.727 ∆𝑃𝑓𝑑𝑒𝑠𝑖𝑔𝑛 [bar] - - 155.4
𝜑 0.0460 0.0411
𝜂𝐿 0.968 0.966 𝐶𝑚 0.005 0.005 0.005
𝐹𝜂̇ 𝐿 𝑑𝑒𝑠𝑖𝑔𝑛 [kmol/hr] 174 188
Table 6. Bypass compressor impeller exit parameters and
∆𝑃𝜂𝐿 𝑑𝑒𝑠𝑖𝑔𝑛 [bar] 44.6 49.0 results for each stage. Fluid states are static values.
𝑓 - 1.031 Stage 1 Stage 2 Stage 3
𝐹𝑓̇ 𝑑𝑒𝑠𝑖𝑔𝑛 [kmol/hr] - 147.0 b2 [m] 0.0062 0.0052 0.0047
∆𝑃𝑓𝑑𝑒𝑠𝑖𝑔𝑛 [bar] - 123.0 𝜑3 𝑠𝑢𝑟𝑔𝑒 0.1824 0.1713 0.1644
𝐶𝑚 0.005 0.005 𝜑3 0.2528 0.2374 0.2279
𝜑3 𝑐ℎ𝑜𝑘𝑒 0.4398 0.4131 0.3964
Table 3. Main compressor impeller exit parameters and kv3 1.37 1.46 1.59
results for each stage. Fluid states are static values. kT3 1.31 1.29 1.28
Stage 1 Stage 2 nv3 1.45 1.57 1.73
b2 [m] 0.0060 0.0059 m3 0.253 0.246 0.239
𝜑3 𝑠𝑢𝑟𝑔𝑒 0.1883 0.1825 V3/Vin 0.8412 0.8694 0.8949
P3 [bar] 109.6 157.7 216.2
𝜑3 0.2611 0.2561
T3 [K] 374.8 410.7 443.8
𝜑3 𝑐ℎ𝑜𝑘𝑒 0.4542 0.4456
ρ3 [kg/m3] 210.9 266.0 319.4
kv3 4.83 5.77
kT3 1.10 1.11

6
For an off-design examination, results are shown for an
increasing temperature at the inlet (MCS1) of the stage 1 main
compressor (MC1) in both a simplified open cycle (Fig. 2) and
closed cycle configuration (Fig. 3), thus allowing a comparison
of the compressor behavior in an application when the inlet
pressure is controlled (constant for the open cycle here) versus
when it is uncontrolled as in the closed cycle. Also, the closed
cycle is more like the final application in the recompression
Brayton cycle.

Figure 4. Stage 1 main compressor results (and total


compressor pressure ratio - 1) for increasing inlet
temperature and constant inlet pressure in the open cycle
Figure 2. Open cycle test configuration

Figure 3. Closed cycle test configuration


For the open cycle, the valve (V1) outlet pressure is fixed at
85.35 bar (the same pressure as the compressor inlet pressure).
For the closed cycle, the compressor inlet pressure changes as
the inlet temperature is increased. For both cycles, the heat
exchanger (Heater) prior to the valve (V1) increases the
Figure 5. Stage 1 main compressor results (and total
temperature to 973°K (heater outlet temperature and turbine inlet
compressor pressure ratio - 1) for increasing inlet
temperature in the 10 MW sCO2 cycle). For the closed cycle, the
temperature and pressure in the closed cycle
heat exchanger (Cooler) prior to the compressor inlet decreases
the inlet temperature to 308°K at design. For the open cycle, the The figure legends are:
inlet temperature to the compressor is a boundary condition, and 𝜑𝑟 : Relative inlet flow coefficient (Value divided by design
for the closed cycle the heat input is varied to obtain the desired value at 308°K). Left axis.
temperature during the test cases. 𝜑3𝑟 : Relative inlet flow coefficient. Left axis.
Select open and closed cycle results are plotted in Figs. 4 I : Impeller work input factor. Left xis.
and 5, respectively, and tabulated in Tables 7 and 8, respectively. CPR – 1 : Compressor pressure ratio across both stages minus 1
Only stage 1 (MC1) results are shown except the pressure ratio (for map plotting). Left axis.
of the total compressor (minus one for plotting purposes, CPR – Pin : Stage 1 inlet pressure in bar. Right axis.
1). The compressor inlet temperature is increased from just Margin-Phi3 : %Surge margin based on exit flow coefficient.
above critical (304°K) to 318°K. The surge margin that would Right axis.
be predicted for the inlet flow coefficient is tabulated and plotted Margin-Phi : %Surge margin based on inlet flow coefficient.
for comparison to the developed exit coefficient. Right axis.
mdot : Mass flow rate through compressor in kg/s. Right axis

7
For the open cycle results presented in Table 7 and Figure increases; however, the total compressor pressure ratio remains
4, th inlet pressure is kept constant. As seen in the table, the nearly constant. The flow coefficients and thus surge margin
polytropic volume coefficient, 𝑛𝜈3 , decreases dramatically as the increase slightly, with the surge margin predictions being nearly
compressor inlet temperature increases. The flow coefficients identical. Figures 4 and 5 are on the same scale and thus the
increase along with the surge margins as the temperature is relative differences can be readily observed. For the impeller
increased. While there is some difference between the margins work input factor and pressure ratio, a directional trend like the
based on the exit vs. the inlet flow coefficients, it does not seem open cycle can be observed. However, the increasing inlet
substantial, and thus in this region of operation, using the pressure is limiting the increase in the volume flow rate
traditional inlet coefficient should not give much error. The (compare 𝜑 in Table 7 and 8) and thus the compressor operating
work input factor, I, and compressor pressure ratio decrease as condition.
would be expected given the increase in inlet flow coefficient The discussion so far has been for a changing inlet
(proportional to volume flow). It is interesting that the mass flow temperature. Taking another example case using the closed
rate decreases by half over the modest temperature change to the cycle, the inlet pressure and temperature are held constant, but
compressor inlet. While not plotted here, note that simulations the valve position after the heater is reduced from 80% to 50%
performed with the bypass compressor, which is in the “safe (Shown as 20% closed to 50% closed in Fig. 6). The results in
zone” discussed by Baltadjiev [7], there is very little change in Fig. 6 show that the pressure ratio and the impeller work
the first stage surge margin for even a 20°K temperature increase coefficient increase as the valve is closed. This is because the
(there is little change in inlet and impeller exit volumetric flow volume flow rate at the inlet and impeller exit both decrease, and
rates also). thus both flow coefficients, with the relative values in Fig. 6
For comparison, simulations were performed for a closed almost on top of each other. Both surge margin predictions
cycle with the inlet temperature controlled by the Cooler (Fig. decrease as expected, with a slight difference. All the
3). Select results are shown in Table 8 and plotted in Fig. 5. As characteristics just discussed are expected.
the inlet temperature is increased, the resulting inlet pressure also
Table 7. Open Cycle – Select first stage results and total compressor
Tin [K] 304 306 308 310 312 314 316 318
First stage
𝜑 0.0434 0.0443 0.0457 0.0487 0.0524 0.0537 0.0543 0.0547
𝜑3 0.2539 0.2561 0.2595 0.2673 0.2821 0.2913 0.2970 0.3009
I 0.7304 0.7278 0.7235 0.7136 0.6950 0.6836 0.6769 0.6722
nv3 9.24 7.36 5.37 3.25 2.11 1.85 1.74 1.67
MU2 0.432 0.469 0.524 0.612 0.662 0.655 0.645 0.635
𝜑 Margin 23.6 25.0 27.4 31.9 36.7 38.2 38.9 39.3
𝜑3 Margin 25.8 26.4 27.4 29.5 33.2 35.4 36.6 37.4
Total compressor
𝑚̇[kg/s] 70.7 68.1 64.4 57.5 47.0 41.1 37.8 35.4
PR 3.05 2.95 2.83 2.58 2.21 2.02 1.91 1.83
Power [kW] 1940 1866 1761 1558 1254 1087 989 920

Table 8. Closed Cycle – Select first stage results and total compressor
Tin [K] 304 306 308 310 312 314 316 318
First stage
𝜑 0.0447 0.0452 0.0458 0.0463 0.0468 0.0472 0.0476 0.0480
𝜑3 0.2546 0.2571 0.2598 0.2621 0.2645 0.2667 0.2692 0.2714
I 0.7302 0.7267 0.7231 0.7200 0.7168 0.7136 0.7105 0.7075
nv3 6.45 5.76 5.26 4.86 4.52 4.24 3.99 3.77
MU2 0.516 0.524 0.528 0.531 0.532 0.533 0.534 0.534
𝜑 Margin 25.7 26.6 27.5 28.3 29.0 29.7 30.3 30.9
𝜑3 Margin 26.0 26.7 27.5 28.1 28.8 29.4 30.0 30.6
Total compressor
𝑚̇[kg/s] 65.9 65.1 64.2 63.4 62.6 61.8 61.04 60.3
PR 2.81 2.82 2.82 2.81 2.81 2.81 2.81 2.81
Power [kW] 1813 1783 1752 1724 1696 1669 1642 1615

8
Figure 6. Closed cycle stage 1 main compressor results Figure 7. Static-to-total density for range of Mach number
(and total compressor pressure ratio - 1) for valve and values of the isentropic volume exponent
decrease
flow coefficient; however, for the range considered the
Baltadjiev [7] indicates that high values and gradients of the difference was not dramatic. The moderate tip Mach number
polytropic or isentropic exponents especially at higher Mach design value is a probable reason for this.
numbers, indicate that using conventional corrections could be For the closed cycle, increasing the compressor inlet
inaccurate and thus more significant differences were expected temperature results in an increasing inlet pressure and reduces
the results presented. It is interesting to note that Eqns. (17) and the inlet density (and thus volumetric flow) change at the
(18) can be manipulated to obtain an expression that looks much compressor inlet compared to the open cycle. Thus, the change
like the isentropic relation for the static-to-total density ratio: in the compressor and overall operating states is also less than
−1
𝜌 𝑘𝑣 − 1 𝑘𝑣 −1
for the open cycle. A closing valve moves the compressor
= (1 + 𝑀𝑎𝑐ℎ#2 ) operating pressure ratio higher and reduces the flow, and the
𝜌𝑡 𝑥
surge margin reduces as expected.
where x = 2 for the isentropic relation. As a rough estimation
9. ACKNOWLEDGEMENTS
from Eqns. (17) and (18) and the main compressor conditions, x
The authors would like to thank the authors in reference [2]
= 3 is used and a range of values for kv is plotted in Fig. 7. Given
for their research and the use of their models as a starting point
the range of Mach values and nv (equal to kv for a polytropic
for the models developed in this work. We would also like to
efficiency of 1) in Table 7, a difference in the density ratio over
thank the Fabio Lambruschini and Dr. Alberto Traverso from the
the range is about 0.06 points (0.89 to 0.95). As can be seen, the
University of Genoa for their cooperative work on sCO2
spread in the density ratio widens for increasing Mach number.
recompression Brayton cycle dynamic modeling [1].
Thus, if the compressor design had started with a higher tip speed
Disclaimer: This report was prepared as an account of work
Mach number (say 0.8 instead of 0.432), perhaps there would
sponsored by an agency of the United States Government.
have been a more significant difference in the exit vs. inlet flow
Neither the United States Government nor any agency thereof,
coefficient results.
nor any of their employees, makes any warranty, express or
implied, or assumes any legal liability or responsibility for the
8. CONCLUSIONS AND OBSERVATIONS
accuracy, completeness, or usefulness of any information,
The results of a multi-stage sCO2 compressor model were
apparatus, product, or process disclosed, or represents that its use
presented for the design of main and bypass compressor for a 10
would not infringe privately owned rights. Reference therein to
MW recompression Brayton cycle. Results were also presented
any specific commercial product, process, or service by trade
where an exit flow coefficient was used as the compressor map
name, trademark, manufacturer, or otherwise does not
parameter for off-design prediction. Testing of the off-design
necessarily constitute or imply its endorsement,
prediction for changes to the inlet temperature was done in an
recommendation, or favoring by the United States Government
open and closed cycle configuration. The results showed that
or any agency thereof. The views and opinions of authors
there was some difference between using the exit vs. the inlet

9
expressed therein do not necessarily state or reflect those of the
United States Government or any agency thereof.

10. REFERENCES
[1] Lambruschini, F., Liese, E., Zitney, S. E., and Traverso, A.,
“Dynamic Model of a 10 MW Supercritical CO2 Recompression
Brayton Cycle”, Proc. of the 2016 ASME Turbo Expo, Seoul, S.
Korea, GT2016-56459.
[2] Modekurti, S., Eslick, J., Omell, B., Bhattacharyya, D.,
Miller, D. C., and Zitney, S. E., “Design, Dynamic Modeling,
and Control of a Multistage CO2 Compression System”, Int. J.
of Greenhouse Gas Control, 2017.
[3] Walsh, P. P., and Fletcher, P., 1998, Gas Turbine
Performance, Blackwell Science/ASME.
[4] Boyce, M. P., “Principles of Operation and Performance
Estimation of Centrifugal Compressors”, Proc. of the 22nd
Turbomachinery Symposium, Dallas, TX, Sep. 14-16, 1993.
[5] Brown, R. N., “Fan Laws, the Use and Limits in Predicting
Centrifugal Compressor Off Design Performance”, Proc. of the
20th Turbomachinery Symposium, Dallas, TX, Sep. 17-19, 1991.
[6] Fink, D. A., “Surge Dynamics and Unsteady Flow
Phenomena”, PhD Dissertation, Dept. of Aeronautics and
Astronautics, Massachusetts Institute of Technology, May,
1988.
[7] Baltadjiev, N. D., 2012, “An Investigation of Real Gas
Effects in Supercritical CO2 Compressors”, Aeronautics and
Astronautics M. S. Thesis for the Mass. Institute of Technology.
[8] Ludtke, K. H., 2004, Process Centrifugal Compressors,
Springer-Verlag, Berlin, ISBN 3-540-40427-9.
[9] Aungier, R. H., 2000, Centrifugal Compressors: A Theory
for Aerodynamic Design and Analysis, ASME Press, NY.
[10] Lemmon, E.W., Huber, M.L., and McLinden, M.O., "NIST
Standard Reference Database 23: NIST Reference Fluid
Thermodynamic and Transport Properties—REFPROP”,
Version 9.1, User's Guide, National Institute of Standards and
Technology, 2013.
[11] Wright, S. A., Radel, F., Conboy, T. M., and Rochau, G. E.,
“Modeling and Experimental Results for Condensing
Supercritical CO2 Power Cycles”, Sandi Report, SAND2010-
8840, January, 2011.

10

You might also like