international journal of refrigeration 33 (2010) 538–552

available at www.sciencedirect.com

w w w . i i fi i r . o r g

journal homepage: www.elsevier.com/locate/ijrefrig

A dynamic model of a vapor compression cycle with

shut-down and start-up operations

Bin Li, Andrew G. Alleyne*

Department of Mechanical Science and Engineering, University of Illinois at Urbana-Champaign, 1206 West Green Street,
MC-244, Urbana, IL 61801, USA

article info abstract

Article history: This paper presents an advanced switched modeling approach for vapor compression cycle
Received 28 December 2008 (VCC) systems used in Air Conditioning and Refrigeration. Building upon recent work
Received in revised form (McKinley and Alleyne, 2008), a complete dynamic VCC model is presented that is able to
5 September 2009 describe the severe transient behaviors in heat exchangers (condenser/evaporator), while
Accepted 7 September 2009 maintaining the moving-boundary framework, under compressor shut-down and start-up
Available online 15 September 2009 operations. The heat exchanger models retain a constant structure, but accommodate
different model representations. Novel switching schemes between different representa-
Keywords: tions and pseudo-state variables are introduced to accommodate the transitions of dynamic
Refrigeration system states in heat exchangers while keeping track of the vapor and liquid refrigerant zones
Compression system during the stop–start transients. Two model validation studies on an experimental system
Régulation show that the complete dynamic model developed in Matlab/Simulink can well predict the
Modelling system dynamics in shut-down and start-up transients.
Simulation ª 2009 Elsevier Ltd and IIR. All rights reserved.
Pressure
Condenser
Evaporator
Heat exchanger
Transient state
Comparison
Experiment

Modèle dynamique d’un cycle à compression de vapeur en

fonctionnement marche/arrêt
Mots clés : Système frigorifique ; Système à compression ; Control ; Modélisation ; Simulation ; Pression ; Condenseur ; Évaporateur ;
Échangeur de chaleur ; Régime transitoire ; Comparaison ; Expérimentation

* Corresponding author. Tel.: þ1 217 244 9993; fax: þ1 217 244 6534.
E-mail address: [email protected] (A.G. Alleyne).
0140-7007/$ – see front matter ª 2009 Elsevier Ltd and IIR. All rights reserved.
doi:10.1016/j.ijrefrig.2009.09.011
 international journal of refrigeration 33 (2010) 538–552 539

Nomenclature Subscripts
A air
Symbols alt alternative
a weight value [dimensionless] c condenser
A area [m2] c1, c2, c3 superheated, two-phase, sub-cooled zone in the
c specific heat [kJ kg1 K1] condenser
D refrigerant-side hydraulic diameter [m] c23 boundary between the two-phase and sub-cooled
f forcing function zone in the condenser
h refrigerant enthalpy [kJ kg1] comp compressor
k refrigerant thermal conductivity [kW m1 K1] ctot complete condensation from saturated vapor to
K gain in the pseudo-state equations; currently set to saturated liquid
five [s1] e evaporator
M refrigerant mass [kg] e1, e2 two-phase, superheated zone in the evaporator
_
m mass flow rate [kg s1] e12 boundary between the two-phase and
€
m time derivative of mass flow rate [kg s2] superheated zone in the evaporator
(mc) thermal capacitance in the heat exchanger etot complete evaporation from saturated liquid to
structure (wall) [kJ K1] saturated vapor
P refrigerant pressure [kPa] f saturated liquid
Q_ heat transfer rate [kW] g saturated vapor
T temperature [K] h enthalpy pseudo-state equation
t time [s] i zone number; for the condenser, i˛{1,2,3}
U overall heat transfer coefficient [kW m2 K1] (1 ¼ superheated, 2 ¼ two-phase, 3 ¼ sub-cooled);
u input for the evaporator, i˛{1,2} (1 ¼ two-phase,
V volume [m3] 2 ¼ superheated)
x state vector iA heat exchanger structure-to-air for zone i
x_ time derivative of the state vector iR heat exchanger structure-to-refrigerant for zone i
Z coefficient matrix in inlet
a refrigerant heat transfer coefficient [kW m2 K1] iwt transported wall temperature across the
g mean void fraction [dimensionless] rightmost boundary of zone i
r refrigerant density [kg m3] min minimum value before switching occurs
s time constant [s] o outlet
z fraction of heat exchanger length covered by zone, pool pool boiling effect
also called normalized zone length PA air at constant pressure
[dimensionless] R refrigerant
Dz normalized length of the liquid–vapor mixture SR refrigerant-to-structure surface
[dimensionless] w heat exchanger structure (wall)

1. Introduction considered. Two heat exchanger modeling approaches are

commonly used in the VCC system modeling: finite-volume
The primary goal of any vapor compression cycle (VCC) distributed-parameter and moving-boundary lumped-
system, such as a refrigeration and/or air-conditioning parameter methods (Bendapudi and Braun, 2002). A validated
system, is to move energy from one physical location to system model of a centrifugal chiller system using the finite-
another. Most VCC systems operate in a mode in which they volume formulation was reported to predict transient
stop and start the refrigerant flow to modulate the amount performance including start-up (Bendapudi et al., 2005). More
of cooling/heating capacity provided to some enclosed recently, Bendapudi et al. (2008) presented a comparative
environmental spaces. Here, this is termed the ‘compressor study of centrifugal chiller system behaviors for start-up and
cycling with shut-down and start-up operations’ or, for short load-change transients with flooded shell-and-tube heat
the ‘stop–start’ problem for dynamic modeling and control exchanger models. In Bendapudi et al. (2008) both finite-
with application to VCC systems. This paper focuses on volume and moving-boundary methods were quantitatively
dynamic modeling of the stop–start operating characteristics compared and the tradeoffs between the two formulations
for control purposes, such as hardware-in-the-loop/soft- were carefully identified. While the moving-boundary
ware-in-the-loop simulation, and embedded system methods were found to be computationally faster, they were
applications. not as computationally robust or accurate as the finite volume
VCC system dynamic modeling is a challenging task in approaches. Considering the target control applications of the
which the balance between complexity and accuracy must be current VCC stop–start system model, we choose to use the
540 international journal of refrigeration 33 (2010) 538–552

moving-boundary approach to model heat exchanger developed by McKinley and Alleyne (2008) is extended to five
dynamics due to its faster speed in real-time computer modes, and the switched evaporator model is developed with
simulation. Furthermore, moving-boundary models provide two modes. The primary difference between the current
industrial practitioners and control engineers with physical efforts and previous model switching approaches lies in the
insights (Rasmussen et al., 2005) into the system’s dynamic choices of the model structure and switching criteria. In the
behaviors that are useful for control design and embedded published approaches (Dhar and Soedel, 1979; Willatzen et al.,
system uses such as diagnostics. 1998; Zhang and Zhang, 2006; Kumar et al., 2008), different
In the moving-boundary modeling approach (He et al., models in heat exchangers were used to accommodate
1998; Rasmussen, 2005; Dhar and Soedel, 1979; Willatzen different operating conditions. These models (Dhar and
et al., 1998; Zhang and Zhang, 2006; Jensen and Tumme- Soedel, 1979; Willatzen et al., 1998; Kumar et al., 2008) may
scheit, 2002; McKinley and Alleyne, 2008; Kumar et al., 2008), experience issues related to transitions between the outputs
heat exchangers are divided into control volumes or zones of different models. Zhang and Zhang (2006) showed that the
based on the fluid phase and model parameters are lumped robustness of the evaporator models was improved through
in each zone. The location of the boundary between zones is time-dependent mean void fraction and continuous values in
a dynamic variable. However, the fixed number of zones in numerical integrations. However, in their work the dynamic
many heat exchanger models severely limits their operating state of pressure was set to be a constant value in simulations
ranges. For example, the compressor cycling with shut-down to achieve this robustness. The present approach uses
and start-up operations produces very large transients which controls techniques from latent tracking algorithms for
result in the destruction and creation of dynamic states (Dhar bumpless transfer (Graebe and Ahlen, 1996) such that the
and Soedel, 1979). The main challenge within this moving- inactive, or unused, states are forced to track some of the
boundary modeling framework becomes how to model active states and the model structure remains unchanged.
transitions in the VCC system between full refrigerant flow, Therefore, all the model’s dynamic state information,
no flow, and back again to full flow condition. Previous work including both active and inactive states, is recorded in
(Dhar and Soedel, 1979; Willatzen et al., 1998; Zhang and transients and the transitions between different model
Zhang, 2006) applied model switching approaches to modes are handled automatically. By applying novel switch-
accommodate multiple modeling frameworks in one simu- ing schemes, this approach also specifically ensures refrig-
lation. A system representation with three condenser models erant mass conservation during mode switching while
(a superheated model, a two-phase model and a sub-cooled keeping tracking of the vapor and liquid refrigerant zones in
model) and one evaporator model was introduced by Dhar the stop–start cycle transients. Finally, given the lack of vali-
and Soedel (1979) to represent the start-up transients. dation in the literature for moving-boundary stop–start
However, there are potential discontinuities and numerical models, model validation is presented for two case studies to
problems with this model switching approach. Willatzen demonstrate the ability to represent compressor shut-down
et al. (1998) described useful techniques for controlling and start-up operating characteristics. To the authors’
inactive states during evaporator model switching, and knowledge, this is the first report in the literature of a stop–
Zhang and Zhang (2006) employed a time-variant mean void start VCC cycle simulated and validated using moving-
fraction in investigating the transient behaviors for the boundary models.
evaporator model under larger disturbances. Neither Will-
atzen et al. (1998) nor Zhang and Zhang (2006) presented
model validation results in their work. Kumar et al. (2008) 2. Switched moving-boundary heat
chose to simulate an air-conditioning system with the exchanger models
compressor stop–start operations by using the evaporator and
condenser model representations in the plant stop–start This section presents switched moving-boundary condenser
steady-state conditions, but this method is limited for and evaporator models. The condenser model (see Fig. 1) is an
understanding the stop–start transient dynamics under extension of the work of McKinley and Alleyne (2008), while
different system operating conditions. Recently, McKinley the evaporator (see Fig. 2) is a new derivation based on
and Alleyne (2008) developed a new switched system model a similar modeling framework. A dynamic stop–start VCC
for a single heat exchanger (condenser). The pseudo-state model is then represented using these models. More infor-
variables were introduced to accommodate the transient mation about the switched heat exchanger modeling
dynamic states, and a novel switching criteria based on void approach and relevant assumptions was discussed in
fraction was presented. The switched model also demon- McKinley and Alleyne (2008).
strated robustness to varying operating conditions while
enabling real-time simulation. 2.1. Extended switched condenser
This paper seeks to develop a dynamic model for the VCC
system to capture the compressor cycling behaviors using the Although the switched moving-boundary condenser model of
switched modeling approach initiated by McKinley and McKinley and Alleyne (2008) accommodates the presence or
Alleyne (2008). The switched modeling framework discussed absence of a sub-cooled region, additional capability is needed
in this paper is applied to develop new switched moving- for complete stop–start transient cycles. Assuming the stop–
boundary heat exchanger models with a constant structure start transition dynamics in terms of refrigerant flow across
that accommodates different model representations or the compressor to be ‘full flow; no flow; full flow again’, the
modes. In particular, the switched condenser model switched condenser model is extended to consist of five
 international journal of refrigeration 33 (2010) 538–552 541

Three-zone
Mode model by
1 McKinley et al.
(2008)
Superheated Two-phase Sub-cooled

Two-zone
Mode model by
2 McKinley et al.
(2008)
Superheated Two-phase

Mode Two-zone
3 model

Two-phase Sub-cooled

Mode One-zone
4 model

Two-phase

Mode One-zone
5 model

Superheated

Fig. 1 – Extended switched moving-boundary condenser model structure.

different model representations or, modes (see Fig. 1). For three zones active already in Fig. 1) to maintain the uniform
a single stop–start cycle, either mode 1 or mode 2 (McKinley state vector xc and provide reasonable initial conditions when
and Alleyne, 2008) becomes the initial and/or final state in the the modes switch from one to another.
condenser model, while the other three modes are interme- T
xc ¼ ½ hc1 Pc hc3 zc1 zc2 Tc1w Tc2w Tc3w gc  (1)
diate conditions during the transients.
The dynamic state vector in Eq. (1) represents the
Zc ðxc ; uc Þx_ c ¼ fc ðxc ; uc Þ (2)
condenser conditions at each instant in time. The uniform
state vector, independent of model representation, enables Air mass flow rate, air inlet temperature and humidity,
the different condenser modes to retain a constant dynamical refrigerant inlet and outlet mass flow rate, and refrigerant
system structure formulated in a nonlinear descriptor form in inlet enthalpy are used as the time varying model inputs,
Eq. (2) (Rasmussen, 2005; Alleyne et al., 2007). Each mode in which can be provided by other component models, such as
the condenser model has its own coefficient matrix Zc (xc, uc) the compressor. The choice of model outputs (e.g. condenser
and forcing function fc (xc, uc) for storing thermodynamic pressure, refrigerant outlet temperature/enthalpy, air outlet
variables, and mass and energy balance terms. The model is temperature) will depend on the interfaces with other system
simulated by calculating x_ c ¼ Z1c fc at each time step and component models. The condenser model mode 1 and mode 2
numerically integrating (with initial conditions) to obtain the were developed by McKinley and Alleyne (2008), in which the
state vector xc. The key to the effectiveness of the approach is air flow and condenser structure governing equations
that the number of states, xc, remains constant. When a mode including structure-to-air and structure-to-refrigerant heat
switch occurs, the inactivated zones merely track the relevant transfer were defined. In this work the same basic governing
dynamic properties of their adjacent zones. An advantage equations will be applied to the other three condenser mode
with the uniform state vector is that it provides consistency in representations (see Fig. 1). Additional details concerning the
numerical simulation for mode switches. The pseudo-state derivations of the refrigerant-side governing equations were
variables, which represent states in inactive zones (McKinley discussed in McKinley and Alleyne (2008), and only the final
and Alleyne, 2008; Willatzen et al., 1998), are implemented in forms of the governing equations for the three modes are
all the condenser model modes (except mode 1 which has all given in this section.
542 international journal of refrigeration 33 (2010) 538–552

Mode Two-zone
1 model

Two-phase Superheated

Mode One-zone
2 model

Two-phase

Fig. 2 – Switched moving-boundary evaporator model structure.

2.1.1. Condenser mode 3: two-phase and sub-cooled two-zone

representation 1 dPc dhc3 hcf  hc3 Q_ c3R
The condenser model is described as a two-zone (two-phase and  þ  _ c23 ¼
m (4)
rc3 dt dt rc3 VcR zc3 rc3 VcR zc3
sub-cooled zones) representation in this mode. A switch could
For the two-phase zone, the conservation equations are:
occur in the condenser model from mode 1 to mode 3 when the
length of superheated zone approaches zero. For the sub-cooled dzc2 zc2 drc2 dPc zc2 drc2 dgc _ c23
m _ cinR
m
zone, the mass and energy conservation equations are: þ þ þ ¼ (5)
dt rc2 dPc dt rc2 dgc dt rc2 VcR rc2 VcR

 
dzc2 zc3 drc3 dhc3 m_ c23 _ coR
m dhc2 1 dPc dhc2 dgc hcf  hc2 Q_ þm_ cinR ðhcin  hc2 Þ
 þ ¼ (3)  þ þ _ c23 ¼ c2R
m
dt rc3 dhc3 dt rc3 VcR rc3 VcR dPc rc2 dt dgc dt rc2 VcR zc2 rc2 VcR zc2
(6)

Refrigerant flow
Oil return flow

Filter/Dryer
Oil Return Line
Oil Separator

Condenser
Fan

Condenser

Evaporator #2
Evaporator
Fan

Compressor

EEV

Evaporator #1
Manual Manual
Valve Valve

Expansion
Valve
Fig. 3 – Schematic of the particular VCC experimental system (shaded portion).
 international journal of refrigeration 33 (2010) 538–552 543

Table 1 – System inputs for both model validation cases.

Inputs Model validation case I Model validation case II

Before Step time for Step time for Before Step time for Step time for
shut-down shut-down start-up shut-down shut-down start-up

Compressor 1000 rpm 1104 s/3111 s 2109 s/4116 s 1500 rpm 527 s 1548 s
speed

Valve open 14% 1104 s/3111 s 2109 s/4116 s 11% No step change No step change
command

Cond. air 0.49 kg s1 No step change No step change 0.3015 kg s1 No step change No step change
mass flow

Evap. air 0.156 kg s1 No step change No step change 0.15 kg s1 No step change No step change
mass flow

Cond. air inlet Time-varying measurement Time-invariant input: 27.942  C

temperature

Evap. air inlet Time-varying measurement Time-invariant input: 26.615  C

temperature

The mean void fraction equation (McKinley and Alleyne, 2.1.2. Condenser mode 4: two-phase one-zone representation
2008) is included for model switching consideration. When the length of the superheated zone in condenser mode 2,
or the length of the sub-cooled zone in mode 3, becomes nearly
dgctot dPc dgc
 ¼ Kc ðgc  gctot Þ (7) zero, the condenser will switch to the two-phase one-zone
dPc dt dt
model given by mode 4 (see Fig. 1). The governing equations for
Since the superheated zone is made inactive in this mode, the refrigerant-side of the active two-phase zone become:
we use: _ cinR  m_ coR
drc2 dPc drc2 dgc m
þ ¼ (14)
dzc1 dPc dt dgc dt zc2 VcR
¼0 (8)
dt
 
dhc2 1 dPc dhc2 dgc Q_ c2R þ m
_ cinR ðhcin  hc2 Þ  m
_ coR ðhco  hc2 Þ
The state equation for the refrigerant enthalpy in the  þ ¼
dPc rc2 dt dgc dt rc2 VcR zc2
superheated zone becomes:
(15)
dhc1 1 dhcin 1 dhcg dPc
¼ þ (9) The constant length of the two-phase zone is described by:
dt 2 dt 2 dPc dt
dzc2
which is derived by differentiating: ¼0 (16)
dt
hcin þ hcg For the inactive superheated zone, Eqs. (8)–(10), (13) still
hc1 ¼ (10)
2 apply, while the pseudo-state Equations (17), (18) are used to
The state equations for the average wall temperature in handle the inactive sub-cooled zone (McKinley and Alleyne,
each control zone (two-phase and sub-cooled zones) are: 2008).

  dhc3

dTc2w 1 Q_ c2R  Q_ c2A dz ¼ Kch hcf  hc3 (17)
¼ þ ðTc2wt  Tc2w Þ c2 (11) dt
dt zc2 ðmcÞcw dt
dTc3w
¼ Kcw ðTc2w  Tc3w Þ (18)
dt
 
dTc3w 1 Q_ c3R  Q_ c3A dz
¼ þ ðTc3w  Tc2wt Þ c2 (12) Equations (8), (9), (11), (13)–(18) ensure that the condenser
dt zc3 ðmcÞcw dt
mode 4 shares the same dynamic states as the other modes as
The pseudo-state equation in Eq. (13) is used to govern the given in Eq. (1).
wall behavior for the inactive superheated zone, which tracks
the wall temperature of the active two-phase zone. 2.1.3. Condenser mode 5: superheated one-zone
representation
dTc1w The condenser model could switch from mode 4 to mode 5
¼ Kcw ðTc2w  Tc1w Þ (13)
dt under certain conditions during shut-down operations as will
Equations (3)–(9) describe the refrigerant characteristics in be seen in the model validation section. Consequently, specific
each control zone, and the wall temperature dynamics are attention is given to the superheated one-zone condenser
presented in Eqs. (11)–(13). These equations can be combined representation. For the inactive zones (two-phase and sub-
to result in the condenser’s mode 3 representation with the cooled zones), the pseudo-state Equations (17)–(19) are applied
uniform dynamic states as given in Eq. (1). to model the wall temperature and enthalpy dynamics.
544 international journal of refrigeration 33 (2010) 538–552

 
dTc1w 1 Q_ c1R  Q_ c1A
Table 2 – Model validation operating conditions. ¼ (23)
dt zc1 ðmcÞcw
Units Model Model
validation validation By combining the Equations (8), (9), (16)–(23), the condenser
case I case II mode 5 model can then be described in the descriptor form of
Eq. (2) using the dynamic states in Eq. (1).
Value Value

Evaporator #2 2.1.4. Switching criteria for the extended switched condenser

Pressure kPa 333.7 241.6
The refrigerant mass conservation is the major concern when
Inlet enthalpy kJ kg1 102.39 96.47
choosing the switching criteria among different model
Refrigerant mass kg s1 0.0063 0.0049
flow rate representations (McKinley and Alleyne, 2008). Given that there
Air inlet temperature 
C 25.46 26.615 are five condenser model modes (see Fig. 1), under different
Air mass flow rate kg s1 0.156 0.15 VCC operating conditions there are many possibilities for the
Air humidity ratio kg kg1 0.0064 0.00864 switching scenarios in the condenser during the stop–start
Compressor cycle dynamics. The switching criteria that are relevant to the
Inlet pressure kPa 333.7 241.6 particular model validation results demonstrated in this paper
Outlet pressure kPa 913.3 975.9 are discussed. Other possible switching conditions between

Inlet temperature C 23.01 25.35 certain condenser modes can be easily incorporated without

Outlet temperature C 54.02 61.29
loss of generality.
Refrigerant mass kg s1 0.0063 0.0049
The switching schemes between mode 1 and mode 2 in the
flow rate
Compressor speed rpm 1000 1500 condenser chosen by McKinley and Alleyne (2008) are used as
guidelines to define other switching criteria. During shut-
Condenser
down operations, possible switches could occur from mode 1
Pressure kPa 913.3 975.9
Refrigerant inlet 
C 48.66 55.69
(three-zone model) to mode 3 (two-phase and sub-cooled two-
temperature zone model), or from mode 2 (superheated and two-phase
Refrigerant mass kg s1 0.0063 0.0049 two-zone model) to mode 4 (two-phase one-zone model) by
flow rate losing the superheated zone. The switching conditions are as

Air inlet temperature C 27.70 27.942 follows:
Air mass flow rate kg s1 0.49 0.3015
Air humidity ratio kg kg1 0.0073 0.00935 zc1 < zcmin (24)
Electric expansion valve and
Inlet pressure kPa 913.3 975.9
Outlet pressure kPa 333.7 241.6 dzc1
Inlet temperature 
C 36.04 32.0 <0 (25)
dt
Refrigerant mass kg s1 0.0063 0.0049
flow rate zcmin is used as a tunable switching threshold (McKinley and
Input signal % 14 11 Alleyne, 2008) indicating the minimum dimensionless length
of the controlled zone such as the superheated zone within
the total condenser length, and is chosen here to be 0.001 as
dTc2w a starting point. Similar criteria can be chosen to trigger the
¼ Kcw ðTc1w  Tc2w Þ (19) switch from condenser mode 3 (two-phase and sub-cooled
dt
two-zone model) to mode 4 (two-phase one-zone model)
An alternate state defined as ‘uniform superheated
based on the normalized length state of the sub-cooled zone.
enthalpy’ (hc1alt) is used, which means that the refrigerant in
The mean void fraction value in the condenser determines
mode 5 has the same enthalpy throughout the condenser.
the switch from mode 4 (two-phase one-zone model) to mode
This differs from the general definition of refrigerant enthalpy
5 (superheated one-zone model), and the vapor volume inside
in a superheated zone (Rasmussen, 2005). The initial condition
the two-phase zone in condenser mode 4 is tracked with
for hc1alt upon initiating mode 5 is the refrigerant saturated
a gain equal to the time derivative of pressure for switching
vapor enthalpy hcg. For the case where the superheated zone is
purposes, as illustrated in Eq. (28).
the only active zone, the mass and energy conservation
equations for the refrigerant are as follows: gc > 1 (26)

drc1 dPc drc1 dhc1alt m_ cinR  m_ coR and

þ ¼ (20)
dPc dt dhc1alt dt zc1 VcR
dgc
1 dPc dhc1alt Q_ c1R þ m
_ cinR ðhcin  hc1alt Þ  m
_ coR ðhco  hc1alt Þ >0 (27)
 þ ¼ (21) dt
rc1 dt dt rc1 VcR zc1
 
The mean void fraction state equation becomes: dzc1alt dPc 
¼ ðz gc  zc1alt Þ (28)
dgc dt dt  c2
¼ Kc ð1  gc Þ (22)
dt For start-up operations, one possible switch in the
The average wall temperature behavior of the superheated condenser happens from mode 5 (superheated one-zone
zone is given as: model) to mode 2 (superheated and two-phase two-zone
 international journal of refrigeration 33 (2010) 538–552 545

Shut-down
Initial state Start-up
Mode 2

Superheated Two-phase

Switched
condenser

Mode 4

Two-phase
End state after
shutdown
Fig. 4 – Switching schemes of different modes in the condenser for model validation case I.

model) when the switching conditions as given in Eqs. cooled two-zone model) are similar to those used for the
(29)–(30) are satisfied. switch from mode 2 (superheated and two-phase two-zone
model) to mode 1 (three-zone model) (McKinley and Alleyne,
Dzc > zcmin (29)
2008).
and Additionally, switching from condenser mode 3 (two-
phase and sub-cooled two-zone model) to mode 1 (three-
dhco zone model), or from condenser mode 4 (two-phase one-zone
<0 (30)
dt model) to mode 2 (superheated and two-phase two-zone
The conditions can be stated as, ‘A prescribed amount of model) occurs when:
liquid–vapor mixture is noticeable and continuing to accu-
mulate in the superheated one-zone condenser model’. The hcin > hcg (32)
normalized length of the liquid–vapor mixture in Eq. (31) is
and
calculated at each time instant by applying the refrigerant
mass conservation principle. dhcin
>0 (33)
r z VcR  McR dt
Dzc ¼ c1alt c1 (31)
rc1alt VcR  rc2alt VcR
This can be explained as, ‘the inlet refrigerant in the
The switching criteria applied to the switch from mode 4 condenser becomes a superheated vapor and the inlet refrig-
(two-phase one-zone model) to mode 3 (two-phase and sub- erant enthalpy is continuing to increase.’

Before
shut-down

A Two-phase Superheated A-A

B
vapor
After
shut-down
liquid
B Two-phase B-B

Fig. 5 – Conceptual representation of refrigerant in the evaporator before and after shut-down.
546 international journal of refrigeration 33 (2010) 538–552

2.2. Switched evaporator The state equations for the average wall temperature of
each control zone are as follows:
In an ideal vapor compression cycle, the inlet refrigerant to  
the evaporator is a saturated liquid–vapor mixture. The dTe1w 1 Q_ e1R  Q_ e1A dz
¼ þ ðTe1wt  Te1w Þ e1 (38)
refrigerant leaving the evaporator ranges from a super- dt ze1 ðmcÞew dt
heated vapor to a saturated liquid–vapor mixture through
 
different VCC operating conditions. Two different represen- dTe2w 1 Q_ e2R  Q_ e2A dz
¼ þ ðTe2w  Te1wt Þ e1 (39)
tations are needed to capture the evaporator dynamics dt ze2 ðmcÞew dt
during the stop–start cycle transients. These are evaporator In the evaporator mode 2 (two-phase one-zone model)
mode 1 (two-phase and superheated two-zone model) and representation, Eq. (39) is replaced with the pseudo-state
evaporator mode 2 (two-phase one-zone model) as illus- Equation (40) to govern the wall temperature in the inactive
trated in Fig. 2. superheated zone by tracking the active two-phase zone state.
The evaporator conditions at each time instant are
described by the dynamic state vector: dTe2w
¼ Kew ðTe1w  Te2w Þ (40)
dt
T
xe ¼ ½ ze1 Pe he2 Te1w Te2w ge  (34)
2.2.1. Evaporator mode 1: two-phase and superheated two-
The descriptor form state derivative Equation (35) is
zone representation
uniform in time for both evaporator model representations,
For the refrigerant-side of the superheated zone, the govern-
but with different coefficient matrices Ze (xe, ue) and forcing
ing equations derived by applying the mass and energy
functions fe (xe, ue) depending on the mode of operation.
Ze ðxe ; ue Þx_ e ¼ fe ðxe ; ue Þ (35) conservation laws are:

Similar model inputs and outputs as those in the switched dze1 ze2 dre2 dPe ze2 dre2 dhe2 m_ e12 _ eoR
m
  þ ¼ (41)
condenser are chosen for the evaporator representations (see dt re2 dPe dt re2 dhe2 dt re2 VeR re2 VeR
Fig. 2). The air flow and evaporator structure governing equa-


tions are derived using the condenser modeling framework 1 dPe dhe2 heg  he2 Q_ m_ eoR he2  heg
 þ  _ e12 ¼ e2R
m (42)
presented in McKinley and Alleyne (2008). The structure-to-air re2 dt dt re2 VeR ze2 re2 VeR ze2
and structure-to-refrigerant heat transfer rates in each control
zone (two-phase and superheated zones) are calculated from: The mass and energy conservation equations for the two-
phase zone are:
Q_ eiA ¼ zei m
_ eA cePA ðTeoiA  TeinA Þ (36)
dze1 ze1 dre1 dPe ze1 dre1 dge m_ e12 _ einR
m
þ þ þ ¼ (43)
Q_ eiR ¼ zei UeiR AeSR ðTeiw  TeiR Þ (37) dt re1 dPe dt re1 dge dt re1 VeR re1 VeR

Data
Condenser Pressure
940 Model
Pressure (kPa)

820

700
100 900 1700 2500 3300 4100 4900

Evaporator Pressure
600
Pressure (kPa)

430

260
100 900 1700 2500 3300 4100 4900
Time (s)

Fig. 6 – Refrigerant pressures for model validation case I.

 international journal of refrigeration 33 (2010) 538–552 547

Condenser air outlet temperature Data

31 Model

Temperature
(C)
29

27
100 900 1700 2500 3300 4100 4900
Evaporator air outlet temperature
28
Temperature
(C)

23

18
100 900 1700 2500 3300 4100 4900
Evaporator refrigerant outlet temperature
29
Temperature
(C)

23

18
100 900 1700 2500 3300 4100 4900
Evaporator superheated temperature
30
Temperature

17
(C)

5
100 900 1700 2500 3300 4100 4900
Time (s)

Fig. 7 – Air outlet and refrigerant temperatures for model validation case I.

Equations (40), (46), the dynamic states of the inactive

  superheated zone are forced to track the corresponding states
dhe1 1 dPe dhe1 dge heg  he1
 þ þ _ e12
m
dPe re1 dt dge dt re1 VeR ze1 of the active zone so as to maintain similarity when the
Q_ þm _ einR ðhein  he1 Þ occasion occurs to switch back to the evaporator mode 1
¼ e1R (44) representation.
re1 VeR ze1
dhe2

¼ Keh heg  he2 (46)
The mean void fraction equation, similar to Eq. (7) in the dt
condenser mode 3, is included in the evaporator model.
The constant length of the active two-phase zone in mode 2
dgetot dPe dge is described by:
 ¼ Ke ðge  getot Þ (45)
dPe dt dt
dze1
¼0 (47)
The combination of equations (38), (39), (41)–(45) results in dt
the evaporator mode 1 representation in Fig. 2 with the For the two-phase zone, the refrigerant-side mass and
dynamic state vector given in Eq. (34). energy conservation equations become:

ze1 dre1 dPe ze1 dre1 dge m_ einR  m_ eoR

2.2.2. Evaporator mode 2: two-phase one-zone representation þ ¼ (48)
re1 dPe dt re1 dge dt re1 VeR
When the refrigerant in the evaporator cannot absorb enough
heat for phase change, less evaporation leads to the disap-  
dhe1 1 dPe dhe1 dge Q_ e1R þ m
_ einR ðhein  he1 Þ  m
_ eoR ðheo  he1 Þ
pearance of the superheated zone. Specifically, a switch could  þ ¼
dPe re1 dt dge dt re1 VeR ze1
occur from evaporator mode 1 (two-phase and superheated
(49)
two-zone model) to mode 2 (two-phase one-zone model). In
the mode 2 representation, the pseudo-state Equation (46) The evaporator mode 2 model in Fig. 2 can then be repre-
causes the superheated zone refrigerant enthalpy to track the sented with Eqs. (38), (40), (46)–(49) in the descriptor form of
saturated vapor enthalpy. In effect, with the pseudo-state Eq. (35) using the uniform dynamic states given in Eq. (34).
548 international journal of refrigeration 33 (2010) 538–552

Initial state
Mode 1 Shut-down
Start-up

Mode 2 Superheated Two-phase Sub-cooled

Mode 2

Superheated Two-phase Switched

Mode 4 condenser

Superheated Two-phase

Two-phase Mode 5

Superheated
End state after
shut-down
Fig. 8 – Switching routines in the condenser for model validation case II.

2.2.3. Switching criteria for the switched evaporator

The switched system approaches for the condenser model
3. Model validation
(McKinley and Alleyne, 2008) are also used as guidelines to
choose the appropriate criteria for switching the evaporator
3.1. Simulation environment
model between these two different modes (see Fig. 2). The
To validate the modeling framework described above, the
conditions to cause the switch from evaporator mode 1 (two-
switched moving-boundary heat exchanger models are
phase and superheated two-zone model) to mode 2 (two-
implemented in Thermosys, a Matlab/Simulink toolbox for
phase one-zone model) depend on the state of the normalized
simulating and analyzing the behaviors of vapor compression
length in the superheated zone, and are defined as:
cycle systems (Alleyne et al., 2007; Rasmussen, 2002). Linear
1  ze1 ¼ ze2 < zemin (50) and nonlinear models for basic components of typical VCC
systems are included in this toolbox. A particular vapor
and
compression cycle system is modeled in Thermosys to simu-
dz dz late stop–start cycle dynamics with compressor shut-down
 e1 ¼ e2 < 0 (51)
dt dt and start-up operations. The system contains a variable speed
The value of zemin is user defined, similar to zcmin in Eq. (24), compressor, an electric expansion valve, the switched heat
and is also set to be 0.001 as a starting point. exchangers (condenser and evaporator), and pipe models
The evaporator mode 2 (two-phase one-zone model) connecting each component. The mass flow rates across the
switches to mode 1 (two-phase and superheated two-zone variable speed compressor and electric expansion valve are
model) when: calculated from empirical performance maps (Rasmussen and
Alleyne, 2004; Eldredge et al., 2008). The inputs to each model
ze1 ðge  getot Þ> zemin (52) are generally the outputs of other component models. For
and instance, the refrigerant inlet and outlet mass flow rates are
the switched evaporator model inputs, yet they themselves
dge are the outputs of the expansion valve and compressor
>0 (53)
dt models, respectively.
The switching condition in Eq. (52) is based on mean void The solution procedure in the condenser model presented
fraction similar to the approach of McKinley and Alleyne by McKinley and Alleyne (2008) is applied in the switched
(2008). The value on the left side of Eq. (52) is the normalized evaporator and condenser model in Thermosys to solve the
length of the excess superheated vapor. This switch can be state derivative equations and linear equation sets. The heat
described as, ‘the existence and further increase of excess transfer coefficients for the refrigerant side and air side are
vapor in the two-phase zone indicates the occurrence of computed separately in the switched heat exchanger
a superheated zone in the evaporator.’ models. The air-to-structure heat transfer coefficients are
 international journal of refrigeration 33 (2010) 538–552 549

Mode 1 2 4 Mode 5 3.2. Experimental system

1

Shut The experimental system used for validation is located at

0.5
down the Air Conditioning and Refrigeration Center at the Univer-
Superheated Zone Length
sity of Illinois at Urbana-Champaign. It is a dual-evaporator
0
‘‘trainer’’ system containing the following components:
1
a tube-and-fin condenser, two tube-and-fin evaporators,
0.5 a suction-line accumulator, a liquid-line receiver, a semi-
hermetic compressor, a variety of expansion devices, and
Two-phase Zone Length
0 a full suite of sensors for temperature, pressure and mass flow
0.2 rate. Manual valves allow the system to be configured in many
different ways. A single evaporator (Evaporator #2) system
0.1 configuration with an electric expansion valve (EEV) in Fig. 3 is
Sub-cooled Zone Length used for the model validation results presented here. The
0 physical parameters and more descriptions about the exper-
500 600 700 800 imental system components are detailed by Rasmussen
Time (s) (2005).
Fig. 9 – Condenser superheated/two-phase/sub-cooled
zone length trace in shut-down transients. 3.3. Model validation results

For the particular vapor compression cycle system (see the

shaded part in Fig. 3), there are four controllable inputs:
calculated by j-factor correlations (Kays and London, 1984). compressor speed, valve opening, condenser air mass flow
The refrigerant-to-structure heat transfer coefficients in the rate and evaporator air mass flow rate. The condenser and
two-phase zone are defined by empirical equations. More evaporator air inlet temperatures are generally regarded as
precisely, the two-phase refrigerant heat transfer correlation the disturbances for this system since there are no actuators
developed by Dobson and Chato (1998) is chosen for the to directly control them. The model validation studies here are
condenser, and the evaporator model uses the correlation divided into two possible cases by the choices of the inputs for
from Wattelet (1994). For the single phase zone (super- compressor shut-down and start-up operations. Model vali-
heated/sub-cooled zone), a Colburn modulus-Reynolds dation case I consists of stop–start steps in both compressor
number correlation (Manglik and Bergles, 1995) is applied for speed and valve opening inputs simultaneously while main-
the refrigerant side heat transfer coefficient calculation. All taining the air flow rates through constant fan speeds. Only
the heat transfer coefficients are updated as the operating stop–start step changes in the compressor speed input are
conditions change during the stop–start cycle simulation, considered in case II. The EEV valve opening and fan speeds
while the refrigerant-to-structure heat transfer coefficient in are held at constant input levels. The steps in the system
the two-phase zone of the evaporator is paid specific inputs for both validation cases are summarized in Table 1
attention to during shut-down transients. More detail on this along with the condenser and evaporator air inlet tempera-
is given subsequently when discussing the model validation ture conditions. The operating conditions before system input
results. changes are shown in Table 2.

Mode 2 Mode 1 3.3.1. Model validation case I: stop–start steps in compressor

1 speed and valve opening
Fig. 4 describes the dynamic switching of different modes
in the condenser model structure for this stop–start cycle
0.6 case. The condenser and evaporator become isolated from
each other when the compressor shuts down and the
Two-phase Zone Length expansion valve is closed. Due to the limits of the two-
0.25
phase refrigerant heat transfer correlation (Wattelet, 1994)
0.75 in the evaporator to handle small, or zero, values of mass
flow rates, a semi-empirical method presented in Eq. (54)
Start-up is employed. Considering the changes to the physical
0.4
nature of the refrigerant state in the evaporator during
shut-down transients, as shown schematically in Fig. 5,
Superheated Zone Length the two-phase refrigerant-to-structure heat transfer coef-
0
ficient is regarded as a weighted function of liquid refrig-
1500 1600 1700 1800
erant heat transfer aepool, related to pool boiling effects
Time (s)
(Incropera and DeWitt, 2002), and saturated vapor refrig-
Fig. 10 – Evaporator two-phase/superheated zone length erant heat transfer aeg. Since the mean void fraction ge is
trace in start-up transients. a clear indication of the amount of refrigerant vapor
550 international journal of refrigeration 33 (2010) 538–552

Condenser Pressure
1200 Data
Model

Pressure (kPa) 900

600
500 1000 1500 2000 2500
Evaporator Pressure
800
Pressure (kPa)

500

200
500 1000 1500 2000 2500
Time (s)

Fig. 11 – Refrigerant pressures for model validation case II.

present in the evaporator during shut-down transients, the 3.3.2. Model validation case II: stop–start steps in
weight a is given as a time-variant exponential function of compressor speed
mean void fraction in the two-phase zone of the evapo- In case II, the expansion valve remains open after the
rator, defined in Eq. (56). compressor shuts down, so the two heat exchangers are never
isolated from one another. This causes equalized refrigerant
ae1 ¼ ð1  aÞaepool þ aaeg (54)
pressures after long durations in shut-down transients.
where aeg is computed from: Considering the slightly higher condenser air inlet tempera-
ture than the evaporator air inlet temperature during
keg shut-down operations (see Table 1), the condenser switches to
aeg ¼ (55)
DeR a mode 5 (superheated one-zone model) representation, and
  the evaporator becomes a mode 2 (two-phase one-zone
t  t0
a ¼ ge þ ðge0  ge Þexp  (56) model) representation with loss of the superheated zone.
s
A semi-empirical heat transfer correlation similar to Eqs. (54)–
where t0 represents the time when the refrigerant flow rate (56) is applied to calculate a two-phase refrigerant heat
entering the evaporator goes to zero, or small, values in transfer coefficient in the evaporator during shut-down
shut-down operations, and ge0 is the mean void fraction at operations. Given the difference of refrigerant mass distribu-
time t0. The time constant s, as a user defined parameter, tions before start-up operations between these two cases, and
reflects the change rate of the two-phase refrigerant heat most of the refrigerant charge in the system is accumulated in
transfer in the evaporator, and is set to be 90 s in this vali- a liquid form inside the evaporator during shut-down tran-
dation case. sients in case II, a first-order compressor mass flow rate
The plots in Figs. 6 and 7 compare experimental data with dynamics with a tunable time constant s in Eq. (57) is incor-
various system model outputs. Sudden changes of evapo- porated in the static compressor to slow the response speed of
rator pressure at the beginning of shut-down or start-up the model during start-up transients. A similar approach was
operations (see Fig. 6) have a sudden influence on the presented in Eldredge et al. (2008). This first order dynamic
superheated temperature in the evaporator presented in represents the redistribution of mass from its stationary
Fig. 7. These transients result from the superheated liquid form inside the evaporator to circulating across the
temperature being the difference between refrigerant outlet compressor and through the overall refrigerant loop.
temperature and saturation temperature based on the
refrigerant pressure. The validation results with two stop– _ comp;static  m
m _ comp
€ comp ¼
m (57)
start cycles in this case demonstrate the capability of the s
switched model approach. where the time constant s is chosen to be 40 s in this case.
 international journal of refrigeration 33 (2010) 538–552 551

Condenser sub-cooled temperature Data

Temperature
8 Model

(C)
4

0
500 1000 1500 2000 2500
Condenser air outlet temperature
Temperature

33
(C)

30

27
500 1000 1500 2000 2500
Evaporator air outlet temperature
Temperature

28
(C)

24

20
500 1000 1500 2000 2500
Time (s)

Fig. 12 – Condenser sub-cooled and air outlet temperatures for model validation case II.

The dynamic switching in the condenser follows the path a reasonable and valuable approach to describe transient
shown in Fig. 8. Fig. 9 describes the condenser mode switch in behaviors for VCC systems.
terms of the length traces of three different zones (super- In light of the gaps between the model and experimental
heated/two-phase/sub-cooled) during shut-down transients, data, further work is needed to quantify their accuracy during
while Fig. 10 presents the evaporator mode switch in start-up the stop–start transients. Error evaluations will depend on the
operations. All zone lengths are dimensionless and normal- important system outputs like heat transfer rates of the
ized to sum to 1.0. Through tracking the amount of vapor refrigerant side and air side during transients. Although an
given in Eq. (28) (see Fig. 9), the switch from condenser mode 4 empirical first-order compressor dynamics was used in the
(two-phase one-zone model) to mode 5 (superheated one- validation, future work will focus on improving the accuracy
zone model) is ensured with refrigerant mass conservation. As of the compressor mass flow rate prediction in start-up tran-
seen in Fig. 10, it takes approximately 30 s to re-gain the sients using a first-principles approach. In addition, building
superheated zone in the evaporator after start-up operations. upon the switched modeling simulation capabilities pre-
The comparisons between experimental data and model sented here, future work will also involve embedded system
outputs are presented in Figs. 11 and 12, which further prove control applications and the development of advanced control
the validity of the switched model method. algorithms for stop–start operations in VCC model systems
interacting with enclosed dynamic environmental spaces.

4. Conclusions Acknowledgments

The authors wish to thank the financial support of the spon-

To the authors’ knowledge, this paper presents the first
soring companies of the Air Conditioning and Refrigeration
moving-boundary dynamic VCC model that captures
Center at the University of Illinois at Urbana-Champaign, and
compressor cycling characteristics with shut-down and start-
also thank Neera Jain and Tom McKinley for providing useful
up operations. Using the modeling framework first presented
suggestions for the manuscript preparations.
in McKinley and Alleyne (2008), the new moving-boundary
heat exchanger models are developed with switched model
representations (e.g. the condenser model has five modes; the references
evaporator model has two modes) to accommodate the
changing numbers of fluid zones during the stop–start tran-
sients. Two experimental model validation cases are given for Alleyne, A.G., Rasmussen, B.P., Keir, M.C., Eldredge, B.D., 2007.
a particular vapor compression cycle system. The results Advances in energy systems modeling and control. In:
demonstrate that the switched heat exchanger modeling is Proceedings of the ACC. IEEE, New York, NY, USA, pp. 4363–4373.
552 international journal of refrigeration 33 (2010) 538–552

Bendapudi, S., Braun, J.E., 2002. A Review of Literature on Kumar, M., Kar, I.N., Ray, A., 2008. State space based modeling
Dynamic Models of Vapor Compression Equipment. ASHRAE and performance evaluation of an air-conditioning system.
Report, No. 4036–5. HVAC&R Res. 14 (5), 797–816.
Bendapudi, S., Braun, J.E., Groll, E.A., 2005. Dynamic model of Manglik, R.M., Bergles, A.E., 1995. Heat transfer and pressure drop
a centrifugal chiller system – model development, numerical correlations for the rectangular offset strip fin compact heat
study, and validation. ASHRAE Trans. 111, 132–148. exchanger. Exp. Therm. Fluid Sci. 10, 171–180.
Bendapudi, S., Braun, J.E., Groll, E.A., 2008. A comparison of McKinley, T.L., Alleyne, A.G., 2008. An advanced nonlinear switched
moving-boundary and finite-volume formulations for heat exchanger model for vapor compression cycles using the
transients in centrifugal chillers. Int. J. Refrigeration 31, moving-boundary method. Int. J. Refrigeration 31, 1253–1264.
1437–1452. Rasmussen, B.P., 2002. Control-oriented modeling of transcritical
Dhar, M., Soedel, W., 1979. Transient analysis of a vapor vapor compression systems. MS thesis, University of Illinois at
compression refrigeration system. In: Proceedings of the 15th Urbana-Champaign.
International Congress of Refrigeration, Venice, vol. 2, pp. Rasmussen, B.P., 2005. Dynamic modeling and advanced control
1031–1067. of air conditioning and refrigeration systems. PhD thesis,
Dobson, M.K., Chato, J.C., 1998. Condensation in smooth University of Illinois at Urbana-Champaign.
horizontal tubes. J. Heat Transfer 120 (1), 193–213. Rasmussen, B.P., Alleyne, A.G., 2004. Control-oriented modeling
Eldredge, B.D., Rasmussen, B.P., Alleyne, A.G., 2008. Moving- of transcritical vapor compression systems. ASME J. Dyn. Syst.
boundary heat exchanger models with variable outlet Meas. Control 126, 54–64.
phase. ASME J. Dyn. Syst. Meas. Control 130 (6) Article No: Rasmussen, B.P., Musser, A., Alleyne, A.G., 2005. Model-driven
061003. system identification of transcritical vapor compression
Graebe, S.F., Ahlen, A., 1996. Dynamic transfer among alternative systems. IEEE Trans. Control Syst. Technol. 13, 444–451.
controllers and its relation to antiwindup controller design. Wattelet, J.P., 1994. Heat Transfer Flow Regimes of Refrigerants in
IEEE Trans. Control Syst. Technol. 4 (1), 92–99. a Horizontal-tube Evaporator. ACRC TR-55. University of
He, X.D., Liu, S., Asada, H., Itoh, H., 1998. Multivariable control of Illinois at Urbana-Champaign.
vapor compression systems. HVAC&R Res. 4 (3), 205–230. Willatzen, M., Pettit, N.B.O.L., Ploug-Sorensen, L., 1998. A general
Incropera, F.P., DeWitt, D.P., 2002. Fundamentals of Heat and dynamic simulation model for evaporators and condensers in
Mass Transfer. John Wiley & Sons, New York, NY, USA. refrigeration. Part I: moving-boundary formulation of two-
Jensen, J.M., Tummescheit, H., 2002. Moving boundary models for phase flows with heat exchange. Int. J. Refrigeration 21 (5),
dynamic simulations of two-phase flows. In: Proceedings of 398–403.
the Second International Modelica Conference, March 18–19, Zhang, W.-J., Zhang, C.-L., 2006. A generalized moving-
Oberpfaffenhofen, Germany, pp. 235–244. boundary model for transient simulation of dry-expansion
Kays, W.M., London, A.L., 1984. Compact Heat Exchangers. evaporators under larger disturbances. Int. J. Refrigeration
McGraw-Hill, Inc., New York, NY, USA. 29, 1119–1127.