Control-oriented modelling for multivariable control of refrigerant dynamics in vapor

ed
compression systems
C. Sanama, X. Xia
Centre of New Energy Systems,

iew
Department of Electrical, Electronic and Computer Engineering,
University of Pretoria,
Private Box X20, Hatfield, 0028, Pretoria, South Africa

Abstract

ev
A multivariable MPC is developed on a VC system to control simultaneously the superheat and
evaporator pressure. A nonlinear control-oriented model is implemented with 12 state variables
and 2 inputs. The nonlinear model is linearized around a carefully selected steady state point so

r
that MPC could be implemented to compute optimal expansion valve opening and compressor
speed with measured feedbacks of superheat and evaporator pressure. The control-oriented model

er
and MPC are implemented on Simulink using 4 interconnected subsystems namely, the VC
system, the MPC system, the reference system and the environment. The coupling effects between
superheat and evaporator pressure are considered. MPC performance was satisfactory in terms of
disturbance rejection and reference tracking whilst considering the coupling effects between
pe
superheat and evaporator pressure.
Keywords: Vapor Compression, MPC, Superheat and Capacity Control
1. Introduction
ot

Dynamic control of vapor compression (VC) systems requires developing simple but accurate
control-oriented dynamic models capturing the complex physical transformation of the circulating
fluid. Advanced dynamic control schemes contribute to improve the performance and efficiency
tn

of VC systems operating across wide range. Dynamic control of VC systems consists in controlling
the expansion valve for superheat regulation and the compressor for capacity regulation [1]. The
superheat regulation was reported by [2] who investigated the stability of the control loop of the
flow through an evaporator with a valve. [3], reported the dynamic behavior of evaporators with
rin

thermostatic expansion valves (TEV’s) with its effects on superheat regulation which could
sometimes experience hunting [4]. [5], suggested a modelling and control strategy to investigate
the dynamic behavior of the refrigerant flow through an evaporator to enhance superheat
regulation. Similar modelling method with mathematical models of TEV’s was adopted by [6] for
ep

investigation of the dynamic behavior of the refrigerant flow through an evaporator. The effects
of sudden variation of the evaporator parameters [7] and the experimental investigation of stable
control systems of an evaporator [8] were suggested for superheat regulation.
Early works on capacity regulation of reciprocating compressors of VC systems were reported by
Pr

[9]. [10], compared the performances of on/off versus capacity regulation of the compressors in
heat pumps and the results demonstrated that capacity control could enhance the efficiency of a

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4182012
 system as the compressor speed could vary to adjust with the load. A complete review of variable

ed
speed capacity control was suggested by [11] which showed the potential in terms of energy
savings of variable speed inverter compressors. [12], compared 3 capacity control schemes namely
hot gas bypass, cylinder unloading and suction gas throttling. The comparison was mainly based
on COP, operating range of temperature and mass fraction of the refrigerant. Development and

iew
implementation of variable loading schemes were suggested by [13] for hydraulic capacity control
of reciprocating compressors. Optimal AC systems control schemes were adopted using capacity
regulation for humid and hot weather conditions [14]. Experimental investigation of stepless
capacity control was adopted on reciprocating compressors using a rotary control valve technology
[15]. The results demonstrated that, significant energy savings with reduced loads and the rotary
valve technology was enabling stepless capacity control continuously. Stepless capacity control

ev
with performance evaluation and optimal schemes was also suggested by [16] with varying
operating conditions. Hot gas bypass was compared to 3 methods namely compressor speed
control, volume clearance and cylinder unloading [17]. The results were useful for the selection of
the adequate capacity regulation scheme based on the design requirements.

r
Superheat and capacity regulation are often controlled independently using single input-single
output (SISO) controller with inherent limitations concerning gain tuning as well as coupling effect
er
between superheat and capacity control [18]. Limitations of SISO schemes are addressable by
adopting multi-input-multi output (MIMO) controllers more robust and performant across wide
operating range. Thermal control of enclosed spaces using VC systems with MIMO controllers
pe
have been reported for the implementation of comprehensive mathematical models [19] and the
development of a robust control strategies with model predictive controller (MPC) [20]. Advanced
mathematical models for thermal control of indoor air conditions have also been introduced by
[21] considering three indoor air parameters namely, temperature, humidity and CO2 level to be
controlled simultaneously with an MPC loop [22] after a carful linearization of the nonlinear
ot

models around a steady operating point taken from experimental results [23]. Similar MPC
strategies could be implemented for dynamic control of VC systems despite its fundamental
difference with thermal control of enclosed spaces. The latter consisting in controlling
tn

simultaneously indoor conditions such as temperature, humidity and CO2 level with the
compressor and evaporator fan speeds [22]. The former on the other hand, performs a simultaneous
superheat and capacity regulation [18] by actuating the expansion valve and compressor.
[18], suggested a control-oriented model with lumped parameters using a moving-boundary
rin

formulation to represent the physical transformation of the refrigerant within the evaporator and
condenser. [24], adopted this model for MIMO control of the refrigerant dynamics in VC systems
using combined linear quadratic Gaussian (LQG) and gain scheduling schemes. [25] and [26],
adopted similar control-oriented model for MIMO control of food process and vehicle air-
ep

conditioning (AC) respectively.

[27] presented recent work on dynamic modelling of VC systems for prediction of the refrigerant
conditions during start-up and shutdown. This work also reviewed recent works on dynamic
Pr

modelling of VC systems such as the nonlinear moving-boundary scheme [28], the component-
based scheme [29] and the modelling of startup and shutdown of VC systems [30]. Furthermore,

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4182012
 [31], compared the moving boundary method with a finite volume scheme for modelling of

ed
dynamic VC systems. The results demonstrated the reliability of the moving boundary scheme for
heat exchanger models considering the phase changes which could also be adopted for modelling
of vehicles’ AC systems [32]. However, the aforementioned works were based on long transient
modelling more effective for prediction during start-up and shutdown rather than for model-based

iew
controller design, which requires short transient for linearization of the model around a steady
operating point [18].
Designing a robust and performant multivariable controller of VC systems requires developing a
nonlinear control-oriented model that could be linearized around a steady operating point.
Furthermore, MPC implementation on multivariable control of indoor air has provided satisfactory

ev
performance over the years in terms of reference tracking and disturbance rejection. Therefore,
MPC implementation for multivariable control of the refrigerant dynamics in VC systems using
control-oriented modelling might be worth investigating as this scheme could provide tremendous
performance in terms of reference tracking and disturbance rejection for superheat and capacity

r
regulation.
The objective work is first, to develop a nonlinear control-oriented model of the refrigerant

er
dynamics in VC systems integrating the refrigerant phase changes in the heat exchangers based on
Navier-Stokes equations enabling the model simulation under wide operating conditions.
Secondly, the linearization of the nonlinear control-oriented model around a steady operating
pe
point. Thirdly, MPC implementation on the linear model for simultaneous superheat and capacity
regulation with consideration of the coupling effects between the two variables.
This work contributes in designing a dynamic multivariable controller for VC systems to control
simultaneously the superheat and capacity regulation whilst considering their coupling effects
using MPC loop on a nonlinear control-oriented model carefully linearized around a steady
ot

operating point.
Corresponding Author - Tel: +27 12 420 2165; Fax: +27 12 362 5000
e-mails: conrad.g@gigs-cm.com , xiaohua.xia@up.ac.za
tn

Nomenclature

𝑚𝑐𝑜𝑚𝑝 Refrigerant mass flow rate in compressor (𝑘𝑔 𝑠 ‒ 1)

rin

ηvol Compressor volumetric efficiency

𝜌𝑐𝑜𝑚𝑝,𝑖𝑛 Refrigerant density at compressor inlet (𝑘𝑔 𝑚 ‒ 3)
𝐷𝑣 ‒1
Compressor displacement volume (𝑚3 𝑠 )
𝑅𝑃𝑀𝑐𝑜𝑚𝑝 Compressor rotational speed (𝑟𝑝𝑚)
𝑚𝑡𝑥𝑣 Mass flow rate in expansion valve (𝑘𝑔 𝑠 ‒ 1)
ep

𝐶𝑡𝑥𝑣 Valve flow coefficient

𝑇𝑆ℎ,𝑜𝑝𝑒𝑟𝑎𝑡𝑖𝑛𝑔 Operating superheat (℃)
𝑇𝑆ℎ,𝑠𝑡𝑎𝑡𝑖𝑐 Static superheat (℃)
𝜌𝑡𝑥𝑣,𝑖𝑛 Refrigerant density at expansion valve inlet (𝑘𝑔 𝑚 ‒ 3)
Pr

𝑃𝑡𝑥𝑣,𝑖𝑛 Refrigerant pressure at expansion valve inlet (𝑏𝑎𝑟)

𝑃𝑡𝑥𝑣,𝑜𝑢𝑡 Refrigerant pressure at expansion valve outlet (𝑏𝑎𝑟)
𝑢 Axial velocity (𝑚 𝑠 ‒ 1)

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4182012
 ℎ Refrigerant enthalpy (𝑘𝐽 𝑘𝑔 ‒ 1)

ed
𝛿𝑄𝑒𝑣𝑎𝑝 Heat absorbed in evaporator (𝑘𝑤)
𝑑𝑒𝑣𝑎𝑝,𝑖𝑛 Evaporator inner diameter (𝑚)
𝛼𝑒𝑣𝑎𝑝 Heat transfer coefficient in evaporator (𝑘𝑤 𝑚 ‒ 2℃ ‒ 1)
𝑇𝑒𝑣𝑎𝑝,𝑤𝑎𝑙𝑙 Wall temperature in evaporator (℃)
𝑇𝑒𝑣𝑎𝑝,𝑟𝑒𝑓 Refrigerant temperature in evaporator (℃)

iew
𝐴𝑒𝑣𝑎𝑝 Evaporator surface (𝑚2)
𝑙𝑒𝑣𝑎𝑝,2𝑝ℎ Two-phase length in evaporator (𝑚)
𝜌𝑒𝑣𝑎𝑝,2𝑝ℎ Density of two-phase refrigerant in evaporator (𝑘𝑔 𝑚 ‒ 3)
𝑃𝑒𝑣𝑎𝑝 Refrigerant pressure in evaporator (𝑏𝑎𝑟)
𝜌𝑒𝑣𝑎𝑝,𝑙𝑖𝑞 Density of refrigerant liquid in evaporator (𝑘𝑔 𝑚 ‒ 3)
𝜌𝑒𝑣𝑎𝑝,𝑔𝑎𝑠 Density of refrigerant gas in evaporator (𝑘𝑔 𝑚 ‒ 3)

ev
(𝑚𝑒𝑣𝑎𝑝,𝑟𝑒𝑓)𝑖𝑛 Refrigerant mass flow rate at evaporator inlet (𝑘𝑔 𝑠 ‒ 1)
(𝑚𝑒𝑣𝑎𝑝,𝑟𝑒𝑓)𝑖𝑛𝑡 Refrigerant mass flow rate at evaporator interface (𝑘𝑔 𝑠 ‒ 1)
𝛾𝑒𝑣𝑎𝑝 Evaporator void fraction
ℎ𝑒𝑣𝑎𝑝,𝑙𝑖𝑞 Enthalpy of liquid refrigerant in evaporator (𝑘𝐽 𝑘𝑔 ‒ 1)

r
ℎ𝑒𝑣𝑎𝑝,𝑔𝑎𝑠 Enthalpy of refrigerant gas in evaporator (𝑘𝐽 𝑘𝑔 ‒ 1)
ℎ𝑒𝑣𝑎𝑝,𝑖𝑛 Refrigerant enthalpy at evaporator inlet (𝑘𝐽 𝑘𝑔 ‒ 1)
ℎ𝑒𝑣𝑎𝑝,𝑖𝑛𝑡
𝑄𝑒𝑣𝑎𝑝,2𝑝ℎ
𝛼𝑒𝑣𝑎𝑝,2𝑝ℎ
er
Refrigerant enthalpy at evaporator interface (𝑘𝐽 𝑘𝑔 ‒ 1)
Heat absorption in evaporator two-phase zone (𝑘𝑤)
Heat transfer coefficient in evaporator two-phase zone (𝑘𝑤 𝑚 ‒ 2℃ ‒ 1
)
pe
𝑇𝑒𝑣𝑎𝑝,𝑤𝑎𝑙𝑙,2𝑝ℎ Wall temperature in evaporator two-phase zone (℃)
𝑇𝑒𝑣𝑎𝑝,𝑟𝑒𝑓,2𝑝ℎ Refrigerant temperature in evaporator two-phase zone (℃)
(𝐶𝜌𝐴)𝑒𝑣𝑎𝑝,𝑤𝑎𝑙𝑙 Thermal capacitance at evaporator wall (𝑘𝐽 ℃ ‒ 1 𝑚 ‒ 1𝑘𝑔)
𝛼𝑒𝑣𝑎𝑝,𝑜𝑢𝑡 Heat transfer coefficient at evaporator outer surface (𝑘𝑤 𝑚 ‒ 2℃ ‒ 1)
𝑑𝑒𝑣𝑎𝑝,𝑜𝑢𝑡 Evaporator outer diameter (𝑚)
ot

𝑇𝑒𝑣𝑎𝑝,𝑤𝑎𝑡𝑒𝑟,𝑖𝑛 Water temperature at evaporator inlet (℃)

𝑙𝑒𝑣𝑎𝑝,𝑠𝑝ℎ Superheat length in evaporator (𝑚)
𝜌𝑒𝑣𝑎𝑝,𝑠𝑝ℎ Density of superheated refrigerant in evaporator (𝑘𝑔 𝑚 ‒ 3)
ℎ𝑒𝑣𝑎𝑝,𝑠𝑝ℎ Enthalpy of superheated refrigerant in evaporator (𝑘𝐽 𝑘𝑔 ‒ 1)
tn

ℎ𝑒𝑣𝑎𝑝,𝑜𝑢𝑡 Refrigerant enthalpy at evaporator outlet (𝑘𝐽 𝑘𝑔 ‒ 1)

(𝑚𝑒𝑣𝑎𝑝,𝑟𝑒𝑓)𝑜𝑢𝑡 Refrigerant mass flow rate at evaporator outlet (𝑘𝑔 𝑠 ‒ 1)
𝑄𝑒𝑣𝑎𝑝,𝑠𝑝ℎ Heat absorption in evaporator superheat zone (𝑘𝑤)
𝛼𝑒𝑣𝑎𝑝,𝑠𝑝ℎ Heat transfer coefficient in evaporator superheat zone (𝑘𝑤 𝑚 ‒ 2℃ ‒ 1)
rin

𝑇𝑒𝑣𝑎𝑝,𝑤𝑎𝑙𝑙,𝑠𝑝ℎ Wall temperature in evaporator superheat zone (℃)

𝑇𝑒𝑣𝑎𝑝,𝑟𝑒𝑓,𝑠𝑝ℎ Refrigerant temperature in evaporator superheat zone (℃)
𝐴𝑐𝑜𝑛𝑑 Condenser surface (𝑚2)
𝑙𝑐𝑜𝑛𝑑,𝑠𝑝ℎ Superheat length in condenser (𝑚)
ep

𝜌𝑐𝑜𝑛𝑑,𝑠𝑝ℎ Density of superheated refrigerant in condenser (𝑘𝑔 𝑚 ‒ 3)

𝑃𝑐𝑜𝑛𝑑 Refrigerant pressure in condenser (𝑘𝑃𝑎)
𝜌𝑐𝑜𝑛𝑑,𝑔𝑎𝑠 Refrigerant density in condenser (𝑘𝑔 𝑚 ‒ 3)
𝑙𝑐𝑜𝑛𝑑,2𝑝ℎ Two-phase length in condenser (𝑚)
(𝑚𝑐𝑜𝑛𝑑,𝑟𝑒𝑓)𝑖𝑛
Pr

Mass flow rate at condenser inlet (𝑘𝑔 𝑠 ‒ 1)

(𝑚𝑐𝑜𝑛𝑑,𝑟𝑒𝑓)𝑖𝑛𝑡,1 Mass flow rate at condenser interface 1 (𝑘𝑔 𝑠 ‒ 1)

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4182012
 ℎ𝑐𝑜𝑛𝑑,𝑠𝑝ℎ Enthalpy of superheated refrigerant in condenser (𝑘𝐽 𝑘𝑔 ‒ 1)

ed
ℎ𝑐𝑜𝑛𝑑,𝑖𝑛𝑡,1 Refrigerant enthalpy at condenser interface 1 (𝑘𝐽 𝑘𝑔 ‒ 1)
ℎ𝑐𝑜𝑛𝑑,𝑖𝑛 Refrigerant enthalpy at condenser inlet (𝑘𝐽 𝑘𝑔 ‒ 1)
𝑄𝑐𝑜𝑛𝑑,𝑠𝑝ℎ Heat absorption in condenser superheat zone (𝑘𝑤)
𝛼𝑐𝑜𝑛𝑑,𝑠𝑝ℎ Heat transfer coefficient in condenser superheat zone (𝑘𝑤 𝑚 ‒ 2℃ ‒ 1)

iew
𝑑𝑐𝑜𝑛𝑑,𝑖𝑛 Condenser inner diameter (𝑚)
𝑇𝑐𝑜𝑛𝑑,𝑤𝑎𝑙𝑙,𝑠𝑝ℎ Wall temperature in condenser superheat zone (℃)
𝑇𝑐𝑜𝑛𝑑,𝑟𝑒𝑓,𝑠𝑝ℎ Refrigerant temperature in condenser superheat zone (℃)
(𝐶𝜌𝐴)𝑐𝑜𝑛𝑑,𝑤𝑎𝑙𝑙 Thermal capacitance at condenser wall (𝑘𝐽 ℃ ‒ 1 𝑚 ‒ 1𝑘𝑔)
𝑑𝑐𝑜𝑛𝑑,𝑜𝑢𝑡 Condenser outer diameter (𝑚)
𝛼𝑐𝑜𝑛𝑑,𝑜𝑢𝑡 Heat transfer coefficient at condenser outer surface (𝑘𝑤 𝑚 ‒ 2℃ ‒ 1)

ev
𝑇𝑐𝑜𝑛𝑑,𝑤𝑎𝑡𝑒𝑟,𝑖𝑛 Water temperature at condenser inlet (℃)
𝜌𝑐𝑜𝑛𝑑,2𝑝ℎ Density of two-phase refrigerant in condenser (𝑘𝑔 𝑚 ‒ 3)
𝜌𝑐𝑜𝑛𝑑,𝑙𝑖𝑞 Density of refrigerant liquid in condenser (𝑘𝑔 𝑚 ‒ 3)
(𝑚𝑐𝑜𝑛𝑑,𝑟𝑒𝑓)𝑖𝑛𝑡,2 Refrigerant mass flow rate at condenser interface 2 (𝑘𝑔 𝑠 ‒ 1)

r
𝛾𝑐𝑜𝑛𝑑 Condenser void fraction
ℎ𝑐𝑜𝑛𝑑,𝑙𝑖𝑞 Enthalpy of liquid refrigerant in condenser (𝑘𝐽 𝑘𝑔 ‒ 1)
ℎ𝑐𝑜𝑛𝑑,𝑔𝑎𝑠
ℎ𝑐𝑜𝑛𝑑,𝑖𝑛𝑡,2
𝑄𝑐𝑜𝑛𝑑,2𝑝ℎ
𝛼𝑐𝑜𝑛𝑑,2𝑝ℎ
er
Enthalpy of refrigerant gas in condenser (𝑘𝐽 𝑘𝑔 ‒ 1)
Refrigerant enthalpy at condenser interface 2 (𝑘𝐽 𝑘𝑔 ‒ 1)
Heat absorption in condenser two-phase zone (𝑘𝑤)
Heat transfer coefficient in condenser two-phase zone (𝑘𝑤 𝑚 ‒ 2℃ ‒ 1)
pe
𝑇𝑐𝑜𝑛𝑑,𝑤𝑎𝑙𝑙,2𝑝ℎ Wall temperature in condenser two-phase zone (℃)
𝑇𝑐𝑜𝑛𝑑,𝑟𝑒𝑓,2𝑝ℎ Refrigerant temperature in condenser two-phase zone (℃)
𝑙𝑐𝑜𝑛𝑑,𝑠𝑢𝑏 Subcooled length in condenser (𝑚)
𝜌𝑐𝑜𝑛𝑑,𝑠𝑢𝑏 Density of subcooled refrigerant in condenser (𝑘𝑔 𝑚 ‒ 3)
(𝑚𝑐𝑜𝑛𝑑,𝑟𝑒𝑓)𝑜𝑢𝑡
ot

Refrigerant mass flow rate at condenser outlet (𝑘𝑔 𝑠 ‒ 1)

ℎ𝑐𝑜𝑛𝑑,𝑠𝑢𝑏 Enthalpy of subcooled refrigerant in condenser (𝑘𝐽 𝑘𝑔 ‒ 1)
ℎ𝑐𝑜𝑛𝑑,𝑜𝑢𝑡 Refrigerant enthalpy at condenser outlet (𝑘𝐽 𝑘𝑔 ‒ 1)
tn

𝑄𝑐𝑜𝑛𝑑,𝑠𝑢𝑏 Heat absorption in condenser subcooled zone (𝑘𝑤)

𝛼𝑐𝑜𝑛𝑑,𝑠𝑢𝑏 Heat transfer coefficient in condenser subcooled zone (𝑘𝑤 𝑚 ‒ 2℃ ‒ 1)
𝑇𝑐𝑜𝑛𝑑,𝑤𝑎𝑙𝑙,𝑠𝑢𝑏 Wall temperature in condenser subcooled zone (℃)
𝑇𝑐𝑜𝑛𝑑,𝑟𝑒𝑓,𝑠𝑢𝑏 Refrigerant temperature in condenser subcooled zone (℃)
rin

2. Control-oriented modelling

Fig. 1 describes a conventional VC system:

ep
Pr

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4182012
 ed
iew
ev
Figure 1 Conventional VC system

r
2.1. Compressor and expansion valve modelling

defined as: er
Assuming adiabatic compression, the refrigerant mass flow rate through the compressor can be

𝑅𝑃𝑀𝑐𝑜𝑚𝑝
pe
𝑚𝑐𝑜𝑚𝑝 = 𝜂𝑣𝑜𝑙𝜌𝑐𝑜𝑚𝑝,𝑖𝑛𝐷𝑣 (1)
60

The mass flow rate through the thermostatic expansion valve can be defined as:

𝑚𝑡𝑥𝑣 = 𝐶𝑡𝑥𝑣(𝑇𝑆ℎ,𝑜𝑝𝑒𝑟𝑎𝑡𝑖𝑛𝑔 ‒ 𝑇𝑆ℎ,𝑠𝑡𝑎𝑡𝑖𝑐) 𝜌𝑡𝑥𝑣,𝑖𝑛(𝑃𝑡𝑥𝑣,𝑖𝑛 ‒ 𝑃𝑡𝑥𝑣,𝑜𝑢𝑡)

ot

(2)

2.2. Heat exchanger modelling

tn

One dimensional refrigerant flow through long, thin and horizontal heat exchanger tubes as well
as negligible heat conduction axially are the assumptions undertaken for simplicity. We consider
cross-flow type of heat exchangers using water as secondary fluid.
rin

2.2.1. Evaporator
ep
Pr

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4182012
 ed
iew
r ev
er
Figure 2 Evaporator model

Neglecting the refrigerant pressure drop due to the refrigerant momentum and viscosity along the
pe
evaporator tubes since it is small compared to the overall pressure head, the equations for
conservation of mass and energy can be written as:

∂𝜌 ∂(𝜌𝑢)
+ =0 (3)
∂𝑡 ∂𝑥
ot

∂(𝜌ℎ) ∂(𝜌𝑢ℎ)
+ + 𝛿𝑄𝑒𝑣𝑎𝑝 = 0 (4)
∂𝑡 ∂𝑥
tn

Where,
4
𝛿𝑄𝑒𝑣𝑎𝑝 = 𝛼 (𝑇 ‒ 𝑇𝑒𝑣𝑎𝑝,𝑟𝑒𝑓)
𝑑𝑒𝑣𝑎𝑝,𝑖𝑛 𝑒𝑣𝑎𝑝 𝑒𝑣𝑎𝑝,𝑤𝑎𝑙𝑙
rin

The refrigerant flow through the evaporator undergoes phase change due to energy gains, therefore
it is divided in two zones namely, the two-phase and superheated zones as illustrated in Fig. 2. The
refrigerant temperature is at saturation and is spatially invariant along the two-phase zone whilst
it increases along the single-phase zone through the evaporator outlet.
ep

2.2.1.1. Two-phase zone

Adopting time-invariant principle on the mean void fraction, the mass conservation along the two-
phase zone could be expressed as:
Pr

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4182012
 𝑑𝜌𝑒𝑣𝑎𝑝,2𝑝ℎ𝑑𝑃𝑒𝑣𝑎𝑝 𝑑𝑙𝑒𝑣𝑎𝑝,2𝑝ℎ (5
+ 𝐴𝑒𝑣𝑎𝑝(𝜌𝑒𝑣𝑎𝑝,𝑙𝑖𝑞 ‒ 𝜌𝑒𝑣𝑎𝑝,𝑔𝑎𝑠) = (𝑚𝑒𝑣𝑎𝑝,𝑟𝑒𝑓)𝑖𝑛 ‒ (𝑚𝑒𝑣𝑎𝑝,𝑟𝑒𝑓)𝑖𝑛𝑡

ed
𝐴𝑒𝑣𝑎𝑝𝑙𝑒𝑣𝑎𝑝,2𝑝ℎ
𝑑𝑃𝑒𝑣𝑎𝑝 𝑑𝑡 𝑑𝑡 )

Where,

𝜌𝑒𝑣𝑎𝑝,2𝑝ℎ = 𝜌𝑒𝑣𝑎𝑝,𝑙𝑖𝑞(1 ‒ 𝛾𝑒𝑣𝑎𝑝) + 𝜌𝑒𝑣𝑎𝑝,𝑔𝑎𝑠𝛾𝑒𝑣𝑎𝑝

iew
The energy conservation along the two-phase zone could be expressed as:

𝑑(𝜌𝑒𝑣𝑎𝑝,𝑙𝑖𝑞ℎ𝑒𝑣𝑎𝑝,𝑙𝑖𝑞) 𝑑(𝜌𝑒𝑣𝑎𝑝,𝑔𝑎𝑠ℎ𝑒𝑣𝑎𝑝,𝑔𝑎𝑠) 𝑑𝑃𝑒𝑣𝑎𝑝

𝐴𝑒𝑣𝑎𝑝𝑙𝑒𝑣𝑎𝑝,2𝑝ℎ ( 𝑑𝑡
(1 ‒ 𝛾𝑒𝑣𝑎𝑝) +
𝑑𝑡
𝛾𝑒𝑣𝑎𝑝 ‒
𝑑𝑡 ) + 𝐴𝑒𝑣𝑎𝑝(1 ‒ 𝛾𝑒𝑣𝑎𝑝)
(6

ev
𝑑𝑙𝑒𝑣𝑎𝑝,2𝑝ℎ
(𝜌𝑒𝑣𝑎𝑝,𝑙𝑖𝑞ℎ𝑒𝑣𝑎𝑝,𝑙𝑖𝑞 ‒ 𝜌𝑒𝑣𝑎𝑝,𝑔𝑎𝑠ℎ𝑒𝑣𝑎𝑝,𝑔𝑎𝑠) )
𝑑𝑡
= (𝑚𝑒𝑣𝑎𝑝,𝑟𝑒𝑓)𝑖𝑛ℎ𝑒𝑣𝑎𝑝,𝑖𝑛 ‒ (𝑚𝑒𝑣𝑎𝑝,𝑟𝑒𝑓)𝑖𝑛𝑡ℎ𝑒𝑣𝑎𝑝,𝑖𝑛𝑡 + 𝑄𝑒𝑣𝑎𝑝,2𝑝ℎ

r
Where,

𝑄𝑒𝑣𝑎𝑝,2𝑝ℎ = 𝛼𝑒𝑣𝑎𝑝,2𝑝ℎ𝜋𝑑𝑒𝑣𝑎𝑝,𝑖𝑛𝑙𝑒𝑣𝑎𝑝,2𝑝ℎ(𝑇𝑒𝑣𝑎𝑝,𝑤𝑎𝑙𝑙,2𝑝ℎ ‒ 𝑇𝑒𝑣𝑎𝑝,𝑟𝑒𝑓,2𝑝ℎ)

er
The wall equation within the two-phase zone could be given as:

𝑑𝑇𝑒𝑣𝑎𝑝,𝑤𝑎𝑙𝑙,2𝑝ℎ
pe
(𝐶𝜌𝐴)𝑒𝑣𝑎𝑝,𝑤𝑎𝑙𝑙
𝑑𝑡 (7)
= 𝛼𝑒𝑣𝑎𝑝,2𝑝ℎ𝜋𝑑𝑒𝑣𝑎𝑝,𝑖𝑛(𝑇𝑒𝑣𝑎𝑝,𝑟𝑒𝑓,2𝑝ℎ ‒ 𝑇𝑒𝑣𝑎𝑝,𝑤𝑎𝑙𝑙,2𝑝ℎ) + 𝛼𝑒𝑣𝑎𝑝,𝑜𝑢𝑡𝜋𝑑𝑒𝑣𝑎𝑝,𝑜𝑢𝑡
(𝑇𝑒𝑣𝑎𝑝,𝑤𝑎𝑡𝑒𝑟,𝑖𝑛 ‒ 𝑇𝑒𝑣𝑎𝑝,𝑤𝑎𝑙𝑙,2𝑝ℎ)

2.2.1.2. Superheat zone

ot

The mass conservation along the superheat zone could be expressed as:

𝑑𝜌𝑒𝑣𝑎𝑝,𝑠𝑝ℎ𝑑𝑃𝑒𝑣𝑎𝑝 𝑑𝑙𝑒𝑣𝑎𝑝,2𝑝ℎ
tn

(8
𝐴𝑒𝑣𝑎𝑝𝑙𝑒𝑣𝑎𝑝,𝑠𝑝ℎ + 𝐴𝑒𝑣𝑎𝑝(𝜌𝑒𝑣𝑎𝑝,𝑔𝑎𝑠 ‒ 𝜌𝑒𝑣𝑎𝑝,𝑠𝑝ℎ) = (𝑚𝑒𝑣𝑎𝑝,𝑟𝑒𝑓)𝑖𝑛𝑡 ‒ (𝑚𝑒𝑣𝑎𝑝,𝑟𝑒𝑓)𝑜𝑢𝑡
𝑑𝑃𝑒𝑣𝑎𝑝 𝑑𝑡 𝑑𝑡 )

The energy conservation along the superheat zone could be expressed as:
rin

𝑑ℎ𝑒𝑣𝑎𝑝,𝑠𝑝ℎ 𝑑𝑃𝑒𝑣𝑎𝑝
(
𝐴𝑒𝑣𝑎𝑝𝑙𝑒𝑣𝑎𝑝,𝑠𝑝ℎ 𝜌𝑒𝑣𝑎𝑝,𝑠𝑝ℎ
𝑑𝑡
‒
𝑑𝑡 )
𝑑𝑙𝑒𝑣𝑎𝑝,2𝑝ℎ ℎ𝑒𝑣𝑎𝑝,𝑜𝑢𝑡 ‒ ℎ𝑒𝑣𝑎𝑝,𝑖𝑛𝑡
(
=‒ (𝑚𝑒𝑣𝑎𝑝,𝑟𝑒𝑓)𝑖𝑛𝑡 ‒ 𝜌𝑒𝑣𝑎𝑝,𝑔𝑎𝑠𝐴𝑒𝑣𝑎𝑝
𝑑𝑡 2 ) (9
‒ (𝑚𝑒𝑣𝑎𝑝,𝑟𝑒𝑓)𝑜𝑢𝑡 )
ep

ℎ𝑒𝑣𝑎𝑝,𝑜𝑢𝑡 ‒ ℎ𝑒𝑣𝑎𝑝,𝑖𝑛𝑡
+ 𝑄𝑒𝑣𝑎𝑝,𝑠𝑝ℎ
2

Where,
Pr

𝑄𝑒𝑣𝑎𝑝,𝑠𝑝ℎ = 𝛼𝑒𝑣𝑎𝑝,𝑠𝑝ℎ𝜋𝑑𝑒𝑣𝑎𝑝,𝑖𝑛𝑙𝑒𝑣𝑎𝑝,𝑠𝑝ℎ(𝑇𝑒𝑣𝑎𝑝,𝑤𝑎𝑙𝑙,𝑠𝑝ℎ ‒ 𝑇𝑒𝑣𝑎𝑝,𝑟𝑒𝑓,𝑠𝑝ℎ)

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4182012
 The wall equation within the superheat zone could be expressed as:

ed
(𝐶𝜌𝐴)𝑒𝑣𝑎𝑝,𝑤𝑎𝑙𝑙( 𝑑𝑇𝑒𝑣𝑎𝑝,𝑤𝑎𝑙𝑙,𝑠𝑝ℎ
𝑑𝑡
+
𝑙𝑒𝑣𝑎𝑝,𝑠𝑝ℎ 𝑑𝑡 )
𝑇𝑒𝑣𝑎𝑝,𝑤𝑎𝑙𝑙,2𝑝ℎ ‒ 𝑇𝑒𝑣𝑎𝑝,𝑤𝑎𝑙𝑙,𝑠𝑝ℎ𝑑𝑙𝑒𝑣𝑎𝑝,2𝑝ℎ
(10
) ‒𝑇
= 𝛼𝑒𝑣𝑎𝑝,𝑠𝑝ℎ𝜋𝑑𝑒𝑣𝑎𝑝,𝑖𝑛(𝑇𝑒𝑣𝑎𝑝,𝑟𝑒𝑓,𝑠𝑝ℎ ‒ 𝑇𝑒𝑣𝑎𝑝,𝑤𝑎𝑙𝑙,𝑠𝑝ℎ) + 𝛼𝑒𝑣𝑎𝑝,𝑜𝑢𝑡𝜋𝑑𝑒𝑣𝑎𝑝,𝑜𝑢𝑡(𝑇𝑒𝑣𝑎𝑝,𝑤𝑎𝑡𝑒𝑟,𝑖𝑛

iew
𝑒𝑣𝑎𝑝,𝑤𝑎𝑙𝑙,𝑠𝑝ℎ)

2.2.2. Condenser

The condenser modelling could be performed with three zones namely, the superheat, two-phase
and subcool zones as illustrated in Fig. 3.

r ev
er
pe
ot
tn

Figure 3 Condenser model

2.2.2.1. Superheat zone

rin

The mass conservation along the superheat zone could be expressed as:

𝑑𝜌𝑐𝑜𝑛𝑑,𝑠𝑝ℎ𝑑𝑃𝑐𝑜𝑛𝑑 𝑑𝑙𝑐𝑜𝑛𝑑,2𝑝ℎ (11

𝐴𝑐𝑜𝑛𝑑𝑙𝑐𝑜𝑛𝑑,𝑠𝑝ℎ + 𝐴𝑐𝑜𝑛𝑑(𝜌𝑐𝑜𝑛𝑑,𝑔𝑎𝑠 ‒ 𝜌𝑐𝑜𝑛𝑑,𝑠𝑝ℎ) = (𝑚𝑐𝑜𝑛𝑑,𝑟𝑒𝑓)𝑖𝑛 ‒ (𝑚𝑐𝑜𝑛𝑑,𝑟𝑒𝑓)𝑖𝑛𝑡,1
𝑑𝑃𝑐𝑜𝑛𝑑 𝑑𝑡 𝑑𝑡 )
ep

The energy conservation along the superheat zone could be expressed as:
Pr

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4182012
 𝑑ℎ𝑐𝑜𝑛𝑑,𝑠𝑝ℎ 𝑑𝑃𝑐𝑜𝑛𝑑
( )

ed
𝐴𝑐𝑜𝑛𝑑𝑙𝑐𝑜𝑛𝑑,𝑠𝑝ℎ 𝜌𝑐𝑜𝑛𝑑,𝑠𝑝ℎ ‒
𝑑𝑡 𝑑𝑡
𝑑𝑙𝑐𝑜𝑛𝑑,2𝑝ℎ ℎ𝑐𝑜𝑛𝑑,𝑖𝑛𝑡,1 ‒ ℎ𝑐𝑜𝑛𝑑,𝑖𝑛
(
=‒ (𝑚𝑐𝑜𝑛𝑑,𝑟𝑒𝑓)𝑖𝑛 ‒ 𝜌𝑐𝑜𝑛𝑑,𝑔𝑎𝑠𝐴𝑐𝑜𝑛𝑑
𝑑𝑡 ) 2
(12
‒ (𝑚𝑐𝑜𝑛𝑑,𝑟𝑒𝑓)𝑖𝑛𝑡,1 )
ℎ𝑐𝑜𝑛𝑑,𝑖𝑛𝑡,1 ‒ ℎ𝑐𝑜𝑛𝑑,𝑖𝑛

iew
+ 𝑄𝑐𝑜𝑛𝑑,𝑠𝑝ℎ
2

Where,
𝑄𝑐𝑜𝑛𝑑,𝑠𝑝ℎ = 𝛼𝑐𝑜𝑛𝑑,𝑠𝑝ℎ𝜋𝑑𝑐𝑜𝑛𝑑,𝑖𝑛𝑙𝑐𝑜𝑛𝑑,𝑠𝑝ℎ(𝑇𝑐𝑜𝑛𝑑,𝑤𝑎𝑙𝑙,𝑠𝑝ℎ ‒ 𝑇𝑐𝑜𝑛𝑑,𝑟𝑒𝑓,𝑠𝑝ℎ)

The wall equation within the superheat zone could be expressed as:

ev
(𝐶𝜌𝐴)𝑐𝑜𝑛𝑑,𝑤𝑎𝑙𝑙 ( 𝑑𝑇𝑐𝑜𝑛𝑑,𝑤𝑎𝑙𝑙,𝑠𝑝ℎ
𝑑𝑡
+
𝑇𝑐𝑜𝑛𝑑,𝑤𝑎𝑙𝑙,2𝑝ℎ ‒ 𝑇𝑐𝑜𝑛𝑑,𝑤𝑎𝑙𝑙,𝑠𝑝ℎ𝑑𝑙𝑐𝑜𝑛𝑑,2𝑝ℎ
𝑙𝑐𝑜𝑛𝑑,𝑠𝑝ℎ 𝑑𝑡 ) (13
) ‒𝑇
= 𝛼𝑐𝑜𝑛𝑑,𝑠𝑝ℎ𝜋𝑑𝑐𝑜𝑛𝑑,𝑖𝑛(𝑇𝑐𝑜𝑛𝑑,𝑟𝑒𝑓,𝑠𝑝ℎ ‒ 𝑇𝑐𝑜𝑛𝑑,𝑤𝑎𝑙𝑙,𝑠𝑝ℎ) + 𝛼𝑐𝑜𝑛𝑑,𝑜𝑢𝑡𝜋𝑑𝑐𝑜𝑛𝑑,𝑜𝑢𝑡(𝑇𝑐𝑜𝑛𝑑,𝑤𝑎𝑡𝑒𝑟,𝑖𝑛

r
𝑐𝑜𝑛𝑑,𝑤𝑎𝑙𝑙,𝑠𝑝ℎ)

2.2.2.2. Two-phase zone

er
Adopting the time-invariant principle on the mean void fraction, the mass conservation along the
two-phase zone could be expressed as:
pe
𝑑𝜌𝑐𝑜𝑛𝑑,2𝑝ℎ𝑑𝑃𝑐𝑜𝑛𝑑 𝑑𝑙𝑐𝑜𝑛𝑑,2𝑝ℎ
𝐴𝑐𝑜𝑛𝑑𝑙𝑐𝑜𝑛𝑑,2𝑝ℎ + 𝐴𝑐𝑜𝑛𝑑(𝜌𝑐𝑜𝑛𝑑,𝑙𝑖𝑞 ‒ 𝜌𝑐𝑜𝑛𝑑,𝑔𝑎𝑠) = (𝑚𝑐𝑜𝑛𝑑,𝑟𝑒𝑓)𝑖𝑛𝑡,1
𝑑𝑃𝑐𝑜𝑛𝑑 𝑑𝑡 𝑑𝑡 (14)
‒ (𝑚𝑒𝑣𝑎𝑝,𝑟𝑒𝑓)𝑖𝑛𝑡,2
ot

Where,

𝜌𝑐𝑜𝑛𝑑,2𝑝ℎ = 𝜌𝑐𝑜𝑛𝑑,𝑙𝑖𝑞(1 ‒ 𝛾𝑐𝑜𝑛𝑑) + 𝜌𝑐𝑜𝑛𝑑,𝑔𝑎𝑠𝛾𝑐𝑜𝑛𝑑

tn

The energy conservation along the two-phase zone could be expressed as:

𝑑(𝜌𝑐𝑜𝑛𝑑,𝑙𝑖𝑞ℎ𝑐𝑜𝑛𝑑,𝑙𝑖𝑞) 𝑑(𝜌𝑐𝑜𝑛𝑑,𝑔𝑎𝑠ℎ𝑐𝑜𝑛𝑑,𝑔𝑎𝑠) 𝑑𝑃𝑐𝑜𝑛𝑑

( (1 ‒ 𝛾𝑐𝑜𝑛𝑑) + ) + 𝐴𝑐𝑜𝑛𝑑(1 ‒ 𝛾𝑐𝑜𝑛𝑑)
rin

𝐴𝑐𝑜𝑛𝑑𝑙𝑐𝑜𝑛𝑑,2𝑝ℎ 𝛾𝑐𝑜𝑛𝑑 ‒
𝑑𝑡 𝑑𝑡 𝑑𝑡
(15
𝑑𝑙𝑐𝑜𝑛𝑑,2𝑝ℎ
(𝜌𝑐𝑜𝑛𝑑,𝑙𝑖𝑞ℎ𝑐𝑜𝑛𝑑,𝑙𝑖𝑞 ‒ 𝜌𝑐𝑜𝑛𝑑,𝑔𝑎𝑠ℎ𝑐𝑜𝑛𝑑,𝑔𝑎𝑠) = (𝑚𝑐𝑜𝑛𝑑,𝑟𝑒𝑓)𝑖𝑛𝑡,1ℎ𝑐𝑜𝑛𝑑,𝑖𝑛𝑡,1 )
𝑑𝑡
‒ (𝑚𝑐𝑜𝑛𝑑,𝑟𝑒𝑓)𝑖𝑛𝑡,2ℎ𝑐𝑜𝑛𝑑,𝑖𝑛𝑡,2 + 𝑄𝑐𝑜𝑛𝑑,2𝑝ℎ
ep

Where,

𝑄𝑐𝑜𝑛𝑑,2𝑝ℎ = 𝛼𝑐𝑜𝑛𝑑,2𝑝ℎ𝜋𝑑𝑐𝑜𝑛𝑑,𝑖𝑛𝑙𝑐𝑜𝑛𝑑,2𝑝ℎ(𝑇𝑐𝑜𝑛𝑑,𝑤𝑎𝑙𝑙,2𝑝ℎ ‒ 𝑇𝑐𝑜𝑛𝑑,𝑟𝑒𝑓,2𝑝ℎ)

Pr

The wall equation within the two-phase zone could be given as:

10

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4182012
 𝑑𝑇𝑐𝑜𝑛𝑑,𝑤𝑎𝑙𝑙,2𝑝ℎ

ed
(𝐶𝜌𝐴)𝑐𝑜𝑛𝑑,𝑤𝑎𝑙𝑙
𝑑𝑡 (16)
= 𝛼𝑐𝑜𝑛𝑑,2𝑝ℎ𝜋𝑑𝑐𝑜𝑛𝑑,𝑖𝑛(𝑇𝑐𝑜𝑛𝑑,𝑟𝑒𝑓,2𝑝ℎ ‒ 𝑇𝑐𝑜𝑛𝑑,𝑤𝑎𝑙𝑙,2𝑝ℎ) + 𝛼𝑐𝑜𝑛𝑑,𝑜𝑢𝑡𝜋𝑑𝑐𝑜𝑛𝑑,𝑜𝑢𝑡
(𝑇𝑐𝑜𝑛𝑑,𝑤𝑎𝑡𝑒𝑟,𝑖𝑛 ‒ 𝑇𝑐𝑜𝑛𝑑,𝑤𝑎𝑙𝑙,2𝑝ℎ)

2.2.2.3. Subcool Zone

iew
The mass conservation along the subcool zone could be expressed as:

𝑑𝜌𝑐𝑜𝑛𝑑,𝑠𝑢𝑏𝑑𝑃𝑐𝑜𝑛𝑑 𝑑𝑙𝑐𝑜𝑛𝑑,2𝑝ℎ (17

𝐴𝑐𝑜𝑛𝑑𝑙𝑐𝑜𝑛𝑑,𝑠𝑢𝑏 + 𝐴𝑐𝑜𝑛𝑑(𝜌𝑐𝑜𝑛𝑑,𝑠𝑢𝑏 ‒ 𝜌𝑐𝑜𝑛𝑑,𝑙𝑖𝑔) = (𝑚𝑐𝑜𝑛𝑑,𝑟𝑒𝑓)𝑖𝑛𝑡,2 ‒ (𝑚𝑐𝑜𝑛𝑑,𝑟𝑒𝑓)𝑜𝑢𝑡
𝑑𝑃𝑐𝑜𝑛𝑑 𝑑𝑡 𝑑𝑡 )

ev
The energy conservation along the subcool zone could be expressed as:
𝑑ℎ𝑐𝑜𝑛𝑑,𝑠𝑢𝑏 𝑑𝑃𝑐𝑜𝑛𝑑
(
𝐴𝑐𝑜𝑛𝑑𝑙𝑐𝑜𝑛𝑑,𝑠𝑢𝑏 𝜌𝑐𝑜𝑛𝑑,𝑠𝑢𝑏
𝑑𝑡
‒
𝑑𝑡 )

r
𝑑𝑙𝑐𝑜𝑛𝑑,𝑠𝑢𝑏 ℎ𝑐𝑜𝑛𝑑,𝑜𝑢𝑡 ‒ ℎ𝑐𝑜𝑛𝑑,𝑖𝑛𝑡,2
(
=‒ (𝑚𝑐𝑜𝑛𝑑,𝑟𝑒𝑓)𝑖𝑛𝑡,2 ‒ 𝜌𝑐𝑜𝑛𝑑,𝑙𝑖𝑞𝐴𝑐𝑜𝑛𝑑
𝑑𝑡 ) 2
(18
‒ (𝑚𝑐𝑜𝑛𝑑,𝑟𝑒𝑓)𝑖𝑛𝑡,2)
ℎ𝑐𝑜𝑛𝑑,𝑜𝑢𝑡 ‒ ℎ𝑐𝑜𝑛𝑑,𝑖𝑛𝑡,2

Where,
2
+ 𝑄𝑐𝑜𝑛𝑑,𝑠𝑢𝑏
er
pe
𝑄𝑐𝑜𝑛𝑑,𝑠𝑢𝑏 = 𝛼𝑐𝑜𝑛𝑑,𝑠𝑢𝑏𝜋𝑑𝑐𝑜𝑛𝑑,𝑖𝑛𝑙𝑐𝑜𝑛𝑑,𝑠𝑢𝑏(𝑇𝑐𝑜𝑛𝑑,𝑤𝑎𝑙𝑙,𝑠𝑢𝑏 ‒ 𝑇𝑐𝑜𝑛𝑑,𝑟𝑒𝑓,𝑠𝑢𝑏)

The wall equation within the subcool zone could be expressed as:

(𝐶𝜌𝐴)𝑐𝑜𝑛𝑑,𝑤𝑎𝑙𝑙( 𝑑𝑇𝑐𝑜𝑛𝑑,𝑤𝑎𝑙𝑙,𝑠𝑢𝑏
𝑑𝑡
+
𝑇𝑐𝑜𝑛𝑑,𝑤𝑎𝑙𝑙,2𝑝ℎ ‒ 𝑇𝑐𝑜𝑛𝑑,𝑤𝑎𝑙𝑙,𝑠𝑢𝑏𝑑𝑙𝑐𝑜𝑛𝑑,2𝑝ℎ
𝑙𝑐𝑜𝑛𝑑,𝑠𝑝ℎ 𝑑𝑡 ) (19
ot

) ‒𝑇
= 𝛼𝑐𝑜𝑛𝑑,𝑠𝑢𝑏𝜋𝑑𝑐𝑜𝑛𝑑,𝑖𝑛(𝑇𝑐𝑜𝑛𝑑,𝑟𝑒𝑓,𝑠𝑢𝑏 ‒ 𝑇𝑐𝑜𝑛𝑑,𝑤𝑎𝑙𝑙,𝑠𝑢𝑏) + 𝛼𝑐𝑜𝑛𝑑,𝑜𝑢𝑡𝜋𝑑𝑐𝑜𝑛𝑑,𝑜𝑢𝑡(𝑇𝑐𝑜𝑛𝑑,𝑤𝑎𝑡𝑒𝑟,𝑖𝑛
𝑐𝑜𝑛𝑑,𝑤𝑎𝑙𝑙,𝑠𝑢𝑏)
tn

3. State space formulation

The state space formulation could be expressed as:

rin

𝐷𝑋 = 𝐹(𝑋,𝑈) + 𝐺(𝑍) (20)

The decoupling matrix 𝐷 could be defined as:

ep
Pr

11

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4182012
 [ ]
𝐷11 𝐷12 0 0 0 0 0 0 0 0 0 0

ed
0 0 𝐷23 0 0 0 0 0 0 0 0 0
𝐷31 𝐷32 0 𝐷34 0 0 0 0 0 0 0 0
0 𝐷42 0 0 𝐷45 0 0 0 0 0 0 0
𝐷51 𝐷52 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 𝐷66 𝐷67 𝐷68 0 0 0 0
𝐷= 0

iew
0 0 0 0 0 𝐷77 0 𝐷79 0 0 0 (21)
0 0 0 0 0 𝐷86 𝐷87 0 0 0 0 0
0 0 0 0 0 0 0 0 0 𝐷910 0 0
0 0 0 0 0 𝐷106 𝐷107 0 0 0 𝐷1011 0
0 0 0 0 0 0 𝐷117 0 0 0 0 𝐷1112
0 0 0 0 0 𝐷126 𝐷126 0 0 0 0 0

ev
The components of 𝐷 are given in Table 1 as follows:
𝑑(𝜌𝑒𝑣𝑎𝑝,𝑙𝑖𝑞ℎ𝑒𝑣𝑎𝑝,𝑙𝑖𝑞) 𝑑(𝜌𝑒𝑣𝑎𝑝,𝑔𝑎𝑠ℎ𝑒𝑣𝑎𝑝,𝑔𝑎𝑠)
( (1 ‒ 𝛾𝑒𝑣𝑎𝑝) + )

r
𝐷11 = 𝐴𝑒𝑣𝑎𝑝𝑙𝑒𝑣𝑎𝑝,2𝑝ℎ 𝛾𝑒𝑣𝑎𝑝 ‒ 1
𝑑𝑃𝑒𝑣𝑎𝑝 𝑑𝑃𝑒𝑣𝑎𝑝
𝐷12 = 𝐴𝑒𝑣𝑎𝑝(1 ‒ 𝛾𝑒𝑣𝑎𝑝)(𝜌𝑒𝑣𝑎𝑝,𝑙𝑖𝑞ℎ𝑒𝑣𝑎𝑝,𝑙𝑖𝑞 ‒ 𝜌𝑒𝑣𝑎𝑝,𝑔𝑎𝑠ℎ𝑒𝑣𝑎𝑝,𝑔𝑎𝑠)

er
𝐷23 = (𝐶𝜌𝐴)𝑒𝑣𝑎𝑝,𝑤𝑎𝑙𝑙
𝐷31 = ‒ 𝐴𝑒𝑣𝑎𝑝𝑙𝑒𝑣𝑎𝑝,𝑠𝑝ℎ

𝐷32 = ‒ 𝜌𝑒𝑣𝑎𝑝,𝑔𝑎𝑠𝐴𝑒𝑣𝑎𝑝
ℎ𝑒𝑣𝑎𝑝,𝑜𝑢𝑡 ‒ ℎ𝑒𝑣𝑎𝑝,𝑖𝑛𝑡
pe
2
𝐷34 = 𝑒𝑣𝑎𝑝 𝑒𝑣𝑎𝑝,𝑠𝑝ℎ 𝑒𝑣𝑎𝑝,𝑠𝑝ℎ
𝐴 𝑙 𝜌
𝑇𝑒𝑣𝑎𝑝,𝑤𝑎𝑙𝑙,2𝑝ℎ ‒ 𝑇𝑒𝑣𝑎𝑝,𝑤𝑎𝑙𝑙,𝑠𝑝ℎ
𝐷42 = (𝐶𝜌𝐴)𝑒𝑣𝑎𝑝,𝑤𝑎𝑙𝑙
𝑙𝑒𝑣𝑎𝑝,𝑠𝑝ℎ
𝐷45 = (𝐶𝜌𝐴)𝑒𝑣𝑎𝑝,𝑤𝑎𝑙𝑙

(
𝑑𝜌𝑒𝑣𝑎𝑝,2𝑝ℎ 𝑑𝜌𝑒𝑣𝑎𝑝,𝑠𝑝ℎ
)
ot

𝐷51 = 𝐴𝑒𝑣𝑎𝑝𝑙𝑒𝑣𝑎𝑝,2𝑝ℎ +
𝑑𝑃𝑒𝑣𝑎𝑝 𝑑𝑃𝑒𝑣𝑎𝑝
𝐷52 = 𝐴𝑒𝑣𝑎𝑝(𝜌𝑒𝑣𝑎𝑝,𝑙𝑖𝑞 ‒ 𝜌𝑒𝑣𝑎𝑝,𝑠𝑝ℎ)
tn

𝐷66 =‒ 𝐴𝑐𝑜𝑛𝑑𝑙𝑐𝑜𝑛𝑑,𝑠𝑝ℎ
ℎ𝑐𝑜𝑛𝑑,𝑖𝑛𝑡,1 ‒ ℎ𝑐𝑜𝑛𝑑,𝑖𝑛
𝐷67 =‒ 𝜌𝑐𝑜𝑛𝑑,𝑔𝑎𝑠𝐴𝑐𝑜𝑛𝑑
2
𝐷68 = 𝐴𝑐𝑜𝑛𝑑𝑙𝑐𝑜𝑛𝑑,𝑠𝑝ℎ𝜌𝑐𝑜𝑛𝑑,𝑠𝑝ℎ
𝑇𝑐𝑜𝑛𝑑,𝑤𝑎𝑙𝑙,2𝑝ℎ ‒ 𝑇𝑐𝑜𝑛𝑑,𝑤𝑎𝑙𝑙,𝑠𝑝ℎ
rin

𝐷77 = (𝐶𝜌𝐴)𝑐𝑜𝑛𝑑,𝑤𝑎𝑙𝑙
𝑙𝑐𝑜𝑛𝑑,𝑠𝑝ℎ
𝐷79 = (𝐶𝜌𝐴)𝑐𝑜𝑛𝑑,𝑤𝑎𝑙𝑙
𝑑(𝜌𝑐𝑜𝑛𝑑,𝑙𝑖𝑞ℎ𝑐𝑜𝑛𝑑,𝑙𝑖𝑞) 𝑑(𝜌𝑐𝑜𝑛𝑑,𝑔𝑎𝑠ℎ𝑐𝑜𝑛𝑑,𝑔𝑎𝑠)
𝐷86 = 𝐴𝑐𝑜𝑛𝑑𝑙𝑐𝑜𝑛𝑑,2𝑝ℎ ( 𝑑𝑃𝑐𝑜𝑛𝑑
(1 ‒ 𝛾𝑐𝑜𝑛𝑑) +
𝑑𝑃𝑐𝑜𝑛𝑑
𝛾𝑐𝑜𝑛𝑑 ‒ 1 )
ep

𝐷87 = 𝐴𝑐𝑜𝑛𝑑(1 ‒ 𝛾𝑐𝑜𝑛𝑑)(𝜌𝑐𝑜𝑛𝑑,𝑙𝑖𝑞ℎ𝑐𝑜𝑛𝑑,𝑙𝑖𝑞 ‒ 𝜌𝑐𝑜𝑛𝑑,𝑔𝑎𝑠ℎ𝑐𝑜𝑛𝑑,𝑔𝑎𝑠)

𝐷910 = (𝐶𝜌𝐴)𝑐𝑜𝑛𝑑,𝑤𝑎𝑙𝑙
𝐷106 =‒ 𝐴𝑐𝑜𝑛𝑑𝑙𝑐𝑜𝑛𝑑,𝑠𝑢𝑏
Pr

ℎ𝑐𝑜𝑛𝑑,𝑜𝑢𝑡 ‒ ℎ𝑐𝑜𝑛𝑑,𝑖𝑛𝑡,2
𝐷107 = ‒ 𝜌𝑐𝑜𝑛𝑑,𝑙𝑖𝑞𝐴𝑐𝑜𝑛𝑑
2

12

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4182012
 𝐷1011 = 𝐴𝑐𝑜𝑛𝑑𝑙𝑐𝑜𝑛𝑑,𝑠𝑢𝑏𝜌𝑐𝑜𝑛𝑑,𝑠𝑢𝑏

ed
𝑇𝑐𝑜𝑛𝑑,𝑤𝑎𝑙𝑙,2𝑝ℎ ‒ 𝑇𝑐𝑜𝑛𝑑,𝑤𝑎𝑙𝑙,𝑠𝑢𝑏
𝐷117 = (𝐶𝜌𝐴)𝑐𝑜𝑛𝑑,𝑤𝑎𝑙𝑙
𝑙𝑐𝑜𝑛𝑑,𝑠𝑝ℎ
𝐷1112 = (𝐶𝜌𝐴)𝑐𝑜𝑛𝑑,𝑤𝑎𝑙𝑙

( 𝑑𝜌𝑐𝑜𝑛𝑑,𝑠𝑝ℎ 𝑑𝜌𝑐𝑜𝑛𝑑,2𝑝ℎ 𝑑𝜌𝑐𝑜𝑛𝑑,𝑠𝑢𝑏

)

iew
𝐷126 = 𝐴𝑐𝑜𝑛𝑑𝑙𝑐𝑜𝑛𝑑,𝑠𝑝ℎ + +
𝑑𝑃𝑐𝑜𝑛𝑑 𝑑𝑃𝑐𝑜𝑛𝑑 𝑑𝑃𝑐𝑜𝑛𝑑
𝐷127 = 𝐴𝑐𝑜𝑛𝑑(𝜌𝑐𝑜𝑛𝑑,𝑠𝑢𝑏 ‒ 𝜌𝑐𝑜𝑛𝑑,𝑠𝑝ℎ)

Table 1 Components of the 𝑫 Matrix

ev
The function 𝐹 could be defined as:
𝑇
𝐹 = [𝐹1 𝐹2 𝐹3 𝐹4 𝐹5 𝐹6 𝐹7 𝐹8 𝐹9 𝐹10 𝐹11 𝐹12] (22)

The components of 𝐹 are given in Table 2 as:

r
𝐹1 = (𝑚𝑒𝑣𝑎𝑝,𝑟𝑒𝑓)𝑖𝑛ℎ𝑒𝑣𝑎𝑝,𝑖𝑛 ‒ (𝑚𝑒𝑣𝑎𝑝,𝑟𝑒𝑓)𝑖𝑛𝑡ℎ𝑒𝑣𝑎𝑝,𝑖𝑛𝑡

er
𝐹2 = 𝛼𝑒𝑣𝑎𝑝,2𝑝ℎ𝜋𝑑𝑒𝑣𝑎𝑝,𝑖𝑛(𝑇𝑒𝑣𝑎𝑝,𝑟𝑒𝑓,2𝑝ℎ ‒ 𝑇𝑒𝑣𝑎𝑝,𝑤𝑎𝑙𝑙,2𝑝ℎ) + 𝛼𝑒𝑣𝑎𝑝,𝑜𝑢𝑡𝜋𝑑𝑒𝑣𝑎𝑝,𝑜𝑢𝑡(𝑇𝑤𝑎𝑡𝑒𝑟,𝑖𝑛 ‒ 𝑇𝑒𝑣𝑎𝑝,𝑤𝑎𝑙𝑙,2𝑝ℎ)

𝐹3 =‒ ((𝑚𝑒𝑣𝑎𝑝,𝑟𝑒𝑓)𝑖𝑛𝑡 + (𝑚𝑒𝑣𝑎𝑝,𝑟𝑒𝑓)𝑜𝑢𝑡)
ℎ𝑒𝑣𝑎𝑝,𝑜𝑢𝑡 ‒ ℎ𝑒𝑣𝑎𝑝,𝑖𝑛𝑡
2
𝐹4 = 𝛼𝑒𝑣𝑎𝑝,𝑠𝑝ℎ𝜋𝑑𝑒𝑣𝑎𝑝,𝑖𝑛(𝑇𝑒𝑣𝑎𝑝,𝑟𝑒𝑓,𝑠𝑝ℎ ‒ 𝑇𝑒𝑣𝑎𝑝,𝑤𝑎𝑙𝑙,𝑠𝑝ℎ) + 𝛼𝑒𝑣𝑎𝑝,𝑜𝑢𝑡𝜋𝑑𝑒𝑣𝑎𝑝,𝑜𝑢𝑡(𝑇𝑤𝑎𝑡𝑒𝑟,𝑖𝑛 ‒ 𝑇𝑒𝑣𝑎𝑝,𝑤𝑎𝑙𝑙,𝑠𝑝ℎ)
pe
𝐹5 = (𝑚𝑒𝑣𝑎𝑝,𝑟𝑒𝑓)𝑖𝑛 ‒ (𝑚𝑒𝑣𝑎𝑝,𝑟𝑒𝑓)𝑜𝑢𝑡
ℎ𝑐𝑜𝑛𝑑,𝑖𝑛𝑡,1 ‒ ℎ𝑐𝑜𝑛𝑑,𝑖𝑛
𝐹6 = ‒ ((𝑚𝑐𝑜𝑛𝑑,𝑟𝑒𝑓)𝑖𝑛 + (𝑚𝑐𝑜𝑛𝑑,𝑟𝑒𝑓)𝑖𝑛𝑡,1)
2
𝐹7 = 𝛼𝑐𝑜𝑛𝑑,𝑠𝑝ℎ𝜋𝑑𝑐𝑜𝑛𝑑,𝑖𝑛(𝑇𝑐𝑜𝑛𝑑,𝑟𝑒𝑓,𝑠𝑝ℎ ‒ 𝑇𝑐𝑜𝑛𝑑,𝑤𝑎𝑙𝑙,𝑠𝑝ℎ) + 𝛼𝑐𝑜𝑛𝑑,𝑜𝑢𝑡𝜋𝑑𝑐𝑜𝑛𝑑,𝑜𝑢𝑡(𝑇𝑤𝑎𝑡𝑒𝑟,𝑖𝑛 ‒ 𝑇𝑐𝑜𝑛𝑑,𝑤𝑎𝑙𝑙,𝑠𝑝ℎ)
𝐹8 = (𝑚𝑐𝑜𝑛𝑑,𝑟𝑒𝑓)𝑖𝑛𝑡,1ℎ𝑐𝑜𝑛𝑑,𝑖𝑛𝑡,1 ‒ (𝑚𝑐𝑜𝑛𝑑,𝑟𝑒𝑓)𝑖𝑛𝑡,2ℎ𝑐𝑜𝑛𝑑,𝑖𝑛𝑡,2
ot

𝐹9 = 𝛼𝑐𝑜𝑛𝑑,2𝑝ℎ𝜋𝑑𝑐𝑜𝑛𝑑,𝑖𝑛(𝑇𝑐𝑜𝑛𝑑,𝑟𝑒𝑓,2𝑝ℎ ‒ 𝑇𝑐𝑜𝑛𝑑,𝑤𝑎𝑙𝑙,2𝑝ℎ) + 𝛼𝑐𝑜𝑛𝑑,𝑜𝑢𝑡𝜋𝑑𝑐𝑜𝑛𝑑,𝑜𝑢𝑡(𝑇𝑤𝑎𝑡𝑒𝑟,𝑖𝑛 ‒ 𝑇𝑐𝑜𝑛𝑑,𝑤𝑎𝑙𝑙,2𝑝ℎ)

ℎ𝑐𝑜𝑛𝑑,𝑜𝑢𝑡 ‒ ℎ𝑐𝑜𝑛𝑑,𝑖𝑛𝑡,2
𝐹10 =‒ ((𝑚𝑐𝑜𝑛𝑑,𝑟𝑒𝑓)𝑖𝑛𝑡,2 + (𝑚𝑐𝑜𝑛𝑑,𝑟𝑒𝑓)𝑖𝑛𝑡,2)
2
tn

𝐹11 = 𝛼𝑐𝑜𝑛𝑑,𝑠𝑝ℎ𝜋𝑑𝑐𝑜𝑛𝑑,𝑖𝑛(𝑇𝑐𝑜𝑛𝑑,𝑟𝑒𝑓,𝑠𝑢𝑏 ‒ 𝑇𝑐𝑜𝑛𝑑,𝑤𝑎𝑙𝑙,𝑠𝑢𝑏) + 𝛼𝑐𝑜𝑛𝑑,𝑜𝑢𝑡𝜋𝑑𝑐𝑜𝑛𝑑,𝑜𝑢𝑡(𝑇𝑤𝑎𝑡𝑒𝑟,𝑖𝑛 ‒ 𝑇𝑐𝑜𝑛𝑑,𝑤𝑎𝑙𝑙,𝑠𝑢𝑏)

𝐹12 = (𝑚𝑐𝑜𝑛𝑑,𝑟𝑒𝑓)𝑖𝑛 ‒ (𝑚𝑐𝑜𝑛𝑑,𝑟𝑒𝑓)𝑜𝑢𝑡

Table 2 Components of the 𝑭 matrix

rin

The state variable 𝑋 could be defined as:

ep
Pr

13

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4182012
 []
𝑃𝑒𝑣𝑎𝑝

ed
𝑙𝑒𝑣𝑎𝑝,2𝑝ℎ
𝑇𝑒𝑣𝑎𝑝,𝑤𝑎𝑙𝑙,2𝑝ℎ
ℎ𝑒𝑣𝑎𝑝,𝑠𝑝ℎ
𝑇𝑒𝑣𝑎𝑝,𝑤𝑎𝑙𝑙,𝑠𝑝ℎ
𝑃𝑐𝑜𝑛𝑑
𝑋= 𝑙

iew
𝑐𝑜𝑛𝑑,2𝑝ℎ (23)
ℎ𝑐𝑜𝑛𝑑,𝑠𝑝ℎ
𝑇𝑐𝑜𝑛𝑑,𝑤𝑎𝑙𝑙,𝑠𝑝ℎ
𝑇𝑐𝑜𝑛𝑑,𝑤𝑎𝑙𝑙,2𝑝ℎ
ℎ𝑐𝑜𝑛𝑑,𝑠𝑢𝑏
𝑇𝑐𝑜𝑛𝑑,𝑤𝑎𝑙𝑙,𝑠𝑢𝑏

ev
The input 𝑈 could be defined as:

[ 𝐶𝑡𝑥𝑣
]

r
𝑈 = 𝑅𝑃𝑀
𝑐𝑜𝑚𝑝
(24)

The function 𝐺 could be defined as:

er
[]
𝑄𝑒𝑣𝑎𝑝,2𝑝ℎ
pe
0
𝑄𝑒𝑣𝑎𝑝,𝑠𝑝ℎ
0
0
𝑄𝑐𝑜𝑛𝑑,𝑠𝑝ℎ
𝐺= 0 (25)
ot

𝑄𝑐𝑜𝑛𝑑,2𝑝ℎ
0
𝑄𝑐𝑜𝑛𝑑,𝑠𝑢𝑏
0
tn

The disturbance 𝑍 could be defined as:

rin

𝑍 = [𝛼𝑒𝑣𝑎𝑝,2𝑝ℎ 𝛼𝑒𝑣𝑎𝑝,𝑠𝑝ℎ 𝛼𝑐𝑜𝑛𝑑,𝑠𝑝ℎ 𝛼𝑐𝑜𝑛𝑑,2𝑝ℎ 𝛼𝑐𝑜𝑛𝑑,𝑠𝑢𝑏]𝑇 (26)

4. MPC implementation
ep

The nonlinear system could be linearized around a steady operating point (𝑋0,𝑈0) using Taylor
first order approximation whilst a close loop MPC could be adopted to reduce the deviations.

{𝑋(𝑘 + 1)𝑌(𝑘)
= 𝐴𝑋(𝑘) + 𝐵𝛥𝑈(𝑘)
= 𝐶𝑋(𝑘) (27)
Pr

14

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4182012
 Where, 𝐴
[ 0𝑇
]

ed
𝐴 = 𝐶𝐴𝑑 𝐼 9 is the state matrix
𝑑 3×3

𝐵𝑑
[ ]
𝐵 = 𝐶𝐵 is the input matrix

iew
𝑑

[
1 0 0 0 0 0 0 0 0 0 0 0
𝐶 = 0 0 0 1 0 0 0 0 0 0 0 0 is the output matrix
0 0 0 0 0 1 0 0 0 0 0 0 ]

ev
With,

∂𝐹
𝐴𝑑 = | 𝑎𝑛𝑑 𝐵𝑑 =
∂𝐹
|

r
∂𝑋 ∂𝑈
𝑋 = 𝑋0, 𝑈 = 𝑈0 𝑋 = 𝑋0, 𝑈 = 𝑈0

𝑌(𝑘𝑖) = 𝐹𝑋(𝑘𝑖) + 𝛷𝛥𝑈(𝑘𝑖)

er
The predicted output vector could be expressed in compact form as:

(28)
pe
Where,

𝑇
𝐹 = [𝐶𝐴 𝐶𝐴2 𝐶𝐴3 ⋯ 𝐶𝐴𝑁𝑝] (29)

And,
ot

[
𝐶𝐵 0 0 ⋯ 0
𝐶𝐴𝐵
2
𝐶𝐵 0 ⋯ 0

] (30)
tn

𝛷 = 𝐶𝐴 𝐵 𝐶𝐴𝐵 𝐶𝐵 ⋯ 0
⋮ ⋮ ⋮ ⋮ ⋮
𝑁𝑝 ‒ 1 𝑁𝑝 ‒ 2
𝐶𝐴 𝐵 𝐶𝐴 𝐵 𝐶𝐴𝑁𝑝 ‒ 2𝐵 ⋯ 𝐶𝐴𝑁𝑝 ‒ 𝑁𝑐𝐵
rin

With the prediction horizon 𝑁𝑝 = 24 and the control horizon 𝑁𝐶 = 2. The data vector containing
the set point information could be expressed as:
ep

𝑅 𝑠 = [1 1 1]𝑇𝑟(𝑘𝑖) (31)

With,

𝑅𝑠 = [1 1 1]𝑇
Pr

15

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4182012
 MPC’s objective is to minimize errors between the output feedbacks and the reference values. This

ed
objective is formulated into a cost function to compute optimal inputs defined as:

12
(32)
𝐽= ∑ [(𝑅 ‒ 𝑌(𝑘 )) (𝑅 ‒ 𝑌(𝑘 )) + 𝛥𝑈(𝑘 ) 𝑅𝛥𝑈(𝑘 )]
𝑠 𝑖
𝑇
𝑠 𝑖 𝑖
𝑇
𝑖

iew
𝑖=1

The first term carries the objective to minimize error whilst the second term considers control
action weighting for large or small error.
Inserting Eq. (28) into Eq. (32) leads to:
12

ev
(33)
𝐽= ∑ [(𝑅 ‒ 𝐹𝑋(𝑘 ) ‒ 𝛷𝛥𝑈(𝑘 )) (𝑅 ‒ 𝐹𝑋(𝑘 ) ‒ 𝛷𝛥𝑈(𝑘 )) + 𝛥𝑈(𝑘 ) 𝑅𝛥𝑈(𝑘 )]
𝑠 𝑖 𝑖
𝑇
𝑠 𝑖 𝑖 𝑖
𝑇
𝑖
𝑖=1

The first derivative of the cost function could be expressed as:

r
12
∂J (34)
∑ [(𝑅 ‒ 𝐹𝑋(𝑘 ) ‒ 𝛷𝛥𝑈(𝑘 )) ( ‒ 𝛷) + 𝛥𝑈(𝑘 ) 𝑅]
∂𝛥𝑈(𝑘𝑖)
=
𝑖=1
𝑠

Setting the derivative equals to zero yields to:

𝑖
er 𝑖
𝑇
𝑖
𝑇
pe
12 12
(35)
(𝛷 𝛷 + 𝑅 )
𝑇
∑ 𝛥𝑈(𝑘 ) 𝑖
𝑇
=𝛷 ∑ (𝑅 ‒ 𝐹𝑋(𝑘 )) 𝑠 𝑖
𝑇

𝑖=1 𝑖=1
Therefore,
ot

12 12
(36)
∑ 𝛥𝑈(𝑘 ) = (𝛷 𝛷 + 𝑅
𝑖
𝑇 )‒1 𝑇 𝛷 ∑ (𝑅 𝑟(𝑘 ) ‒ 𝐹𝑋(𝑘 ))
𝑠 𝑖 𝑖
𝑖=1 𝑖=1
tn

5. Results

MPC was tested on Simulink using MPC toolbox. The control-oriented model was implemented
on Matlab, then linearized around a steady operating point to obtain a Simulink control design
rin

framework. The linear model was used with MPC toolbox for performance evaluation. The
computed variables in Simulink are presented in Fig. 4 as follows:
ep
Pr

16

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4182012
 ed
iew
r ev
er
pe
Figure 4 Computed variables on Simulink

Two inputs were adopted namely, the compressor speed and the expansion valve opening to
control two outputs simultaneously namely the superheat and evaporator pressure (capacity
control). Fig. 5 describes the MPC adopted to compute the optimal inputs with measured output
feedbacks.
ot
tn
rin
ep

Figure 5 MPC model

The time step was 30 s with the entire simulation lasting 3000 s. Fig. 6 summarized the Simulink
Pr

models adopted for the expansion valve (Fig. (6a)), compressor (Fig. (6b)), evaporator (Fig. (6c)),
and condenser (Fig. (6d)).

17

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4182012
 (b)

ed
Compressor

(a)
Expansion Valve

iew
(c) (d)
Evaporator Condenser

r ev
er
Figure 6 Simulink models of VC components

The expansion valve coefficient and the mass flow rate across the expansion valve were carefully
pe
monitored (Fig. (6a)) and so were the compressor speed and the mass flow rate across the
compressor (Fig. (6b)). The refrigerant conditions, the wall temperature and the heat flux were
also monitored through the evaporator model (Fig. (6c)) and the condenser model (Fig. (6d)).
Disturbances affecting the evaporator’s outputs are measured and plotted in Fig. 7 along with the
ot

mass flow rate responses across the expansion valve and compressor.
tn
rin
ep
Pr

18

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4182012
 ed
iew
r ev
Figure 7 System responses to disturbances

er
Temperature, heat, enthalpy and pressure across the evaporator were measured during the entire
simulation. Over 40 parameters were required to test the performance of the MPC controllers.
Some of the parameters such as the geometrical parameters in Table 3 were identical to the
pe
experiment of [27]. The physical properties of the refrigerant were estimated around steady
operating point using empirical equations and thermodynamic tables.
Convection coefficient 20 𝑘𝑤 𝑚 ‒ 2 Evaporator length 3𝑚
‒1
℃
Initial pressure 3.0 𝑏𝑎𝑟 Fin convection coefficient 150 𝑘𝑤 𝑚 ‒ 2℃ ‒ 1
ot

Initial quality 5% Pipe diameter 0.01 𝑚

Controller time constant 20 Pipe thickness 0.0005 𝑚
Displacement volume 0.004 Internal surface area 4.5 𝑚2
tn

Volumetric efficiency 0.8 External surface area 6 𝑚2

Min throat area 0.1 𝑚𝑚2 Foam thickness 0.03 𝑚
Max throat area 1.5 𝑚𝑚2 Copper density 8940 𝑘𝑔 𝑚 ‒ 3
Min throat temperature ‒ 23.5℃ Copper specific heat 390 𝑘𝐽 𝑘𝑔 ‒ 1℃ ‒ 1
Max throat temperature Copper conductivity 400 𝑘𝑤 𝑚 ‒ 1℃ ‒ 1
rin

6.5℃
Condenser length 3𝑚 Foam conductivity 0.03 𝑘𝑤 𝑚 ‒ 1℃ ‒ 1

Table 3 Integrated parameters

ep

The steady operating point adopted for linearization was found from the experiment of [27] where
for cooling operation, superheat is 4oC and evaporator pressure is 3.35 bar. The parameters adopted
for performance evaluation of the control system are listed in Table 3.
MPC performance was evaluated following the disturbance rejection criterion (Fig. 7) with the
Pr

refrigerant flow rates through the expansion valve and compressor adjusting to the disturbances

19

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4182012
 and the set point tracking criterion (Fig. 8) with MPC tracking the reference points of the superheat

ed
and evaporator pressure.
The refrigerant flow rate through the expansion valve and the compressor increased with
increasing environment temperature and absorbed heat. As the environment temperature varied
from 8oC to 18oC for a corresponding from t=750 s to t=2250 s (Fig. (7a)) one could observe the

iew
mass flow rates through the expansion valve (Fig. (7c)) and the compressor (Fig. (7d)) varied
simultaneously from 0.008 𝑘𝑔 𝑠 ‒ 1 to 0.011 𝑘𝑔 𝑠 ‒ 1.
Similarly, as heat is absorbed through the evaporator (Fig. (7b)) one could observe a simultaneous
variation of the mass flow rates through the expansion valve (Fig. (7c)) to a peak of 0.025 𝑘𝑔 𝑠 ‒ 1
and the compressor to a value of 0.041 𝑘𝑔 𝑠 ‒ 1 (Fig. (7d)).

ev
Mass flow rates through the expansion valve and the compressor adjust simultaneously to
overcome sudden disturbances namely environment temperature (Fig. (7a)) and absorbed heat
through the evaporator (Fig. (7b)).

r
er
pe
ot
tn
rin

Figure 8 Coupled reference tracking

Mass flow rate variations through the expansion valve (Fig. (7c)) and compressor (Fig. (7d)) due
ep

to sudden environment temperature (Fig. (7a)) and heat absorbed (Fig. (7b)) changes make the
compressor speed (Fig. (8a)) and expansion valve (Fig. (8c)) vary simultaneously to track the
reference points of the evaporator pressure (Fig. (8b)) and superheat (Fig. (8d)).
Simultaneous superheat and capacity or evaporator pressure control using MPC enhances the
Pr

system performance as the refrigerant flow rate adjusts continuously to track the reference points
through adequate control of the compressor speed and the expansion valve opening. Valve opening

20

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4182012
 and compressor speed rebalance the refrigerant temperature and charge in the system to enhance

ed
the VC system’s performance.
6. Discussion

The control-oriented model was established using component equations similar to [27] however

iew
the refrigerant’s dynamics through the heat exchangers were carefully analyzed by considering the
two-phase and superheat zones within the evaporator in one hand whilst considering the superheat,
two-phase and subcool zone within the condenser on the other hand. The heat transfer coefficients
(HTC’s) within the evaporator and condenser were determined with the correlation of [33] within
the superheat and subcool zones whilst the correlation of [34] was adopted within the two-phase

ev
zones. The water’s HTC was determined using Colburn J-factor.
Correlations and parameters from property tables created issues encountered while computing
HTC’s as they varied at each time step with varying enthalpy, pressure and flow rate. The
numerical discontinuity was addressed using integrated smooth functions enabling fast output

r
tracking.

er
The refrigerant properties were defined from existing R-134a property tables and integrated to the
Matlab code used with the Simulink toolbox which was made of 4 interconnected subsystems
namely, the VC system, the MPC system, the reference system, and the environment.
pe
The control objective was to control simultaneously superheat and evaporator pressure using two
manipulated control variables namely compressor speed and expansion valve opening. The control
system was therefore defined with two inputs and two outputs.
The nonlinear control-oriented model was first linearized around a steady state point carefully
selected using experiments. MPC was then adopted to control superheat and evaporator pressure
ot

with optimal compressor speed and expansion valve opening. MPC received output feedback
values namely superheat and evaporator pressure along with their reference values to calculate
optimal compressor speed and expansion valve coefficient.
tn

Environment temperature and absorbed heat were adopted as disturbances to check the controller
robustness. Therefore, the effect of varying environment temperature and absorbed heat on
superheat and evaporator pressure’s feedbacks were tested on the MPC.
rin

The cost function was formulated to determine the optimal expansion valve coefficient and
compressor speed using a constraint optimizer with measured superheat and evaporator pressure
feedbacks. The difference between measured feedbacks and reference values defined the error
tracking within the prediction horizon whilst the difference between the current and former control
ep

inputs defined the inputs’ variation within the control horizon.

Scaling factors with nominal values were adopted to predict superheat and evaporator pressure
with consistent scale avoiding therefore scaling related issues. Only the first element of the
computed input vectors was adopted for actuation at each time step to avoid instability. Measured
Pr

state variables were repeatedly updated within the prediction horizon at each time step to compute
the optimal input vectors until reaching the reference values.

21

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4182012
 The dynamic formulation of the VC system (Eq. 20) led to 12 state variables but only the feedbacks

ed
of the evaporator’s pressure and superheat were measured by the MPC. Refrigerant conditions
such as temperature, heat, enthalpy and pressure within the evaporator and condenser were also
measured in Simulink (Fig. 4) and could be adopted for other control purposes such as controlling
the condensing pressure and subcooling. In this work, MPC was only adopted to compute optimal

iew
compressor speed and expansion valve opening to track superheat and evaporator pressure’s
reference values considering the coupling effects and the disturbances.
Normalized internal energy from the refrigerant p-h chart was adopted for evaluation of the fluid
characteristics during the simulation so that the refrigerant temperature could be determined
knowing its pressure and internal energy.

ev
The control objective in Fig. 8 is to track the reference values of 4oC for superheat (Fig. (8b)) and
3.35 bar for evaporator pressure (Fig. (8d)) using optimal compressor speed (Fig. (8a)) and
expansion valve coefficient (Fig. (8c)) during the entire simulation which started with the initial
environment temperature and heat absorbed respectively at 8oC (Fig. (7a)) and 0 kw (Fig. (7b)).

r
At t=200 s heat is absorbed (Fig. (7b)) and the refrigerant mass flow rate through the expansion

er
valve (Fig. (7c)) and the compressor (Fig. (7d)) automatically increased to allow more refrigerant
to circulate through the evaporator tubes so that heat could be absorbed rapidly preventing
therefore the output responses from hunting making the MPC robust to sudden disturbances.
pe
In the presence of disturbances, the mass flow rate variation through the expansion valve (Fig.
(7c)) and the compressor (Fig. (7d)) are respectively linked to the expansion valve opening (Fig.
6c) and the compressor speed (Fig. (8a)) therefore, disturbance rejection could be tested and
validated with Fig. 7&8.
Simultaneous control of superheat (Fig. (8d)) and evaporating pressure (Fig. (8b)) could be
ot

conflicting due to the coupling effect between the two variables. The MPC could generate dire
performance without a careful consideration of the coupling effects between the two outputs and
also between the compressor speed and expansion valve opening. To illustrate this, at t=750 s to
tn

t=2250 s the environment temperature rose by 10oC causing automatically the mass flow rate
through the expansion valve (Fig. (7c)) and the compressor (Fig. (7d)) to increase however, only
the compressor speed varied (Fig. (8a)) during that period as the expansion valve opening (Fig.
(8c)) stabilize rapidly so that superheat target could be maintained without a deviation of the
rin

evaporator pressure from its reference value.

7. Conclusion

A nonlinear control-oriented model of a VC system has been formulated with 12 state variables
ep

and 2 inputs. The Nonlinear model was implemented on Simulink using R-134a tables and physical
parameters from experimental work. The nonlinear model was then linearized around a steady
state point to enable MPC implementation for dynamic multivariable control. The control objective
was defined to regulate simultaneously superheat and evaporator pressure. MPC was implemented
Pr

on Simulink so that the superheat and evaporator pressure’s feedbacks could be measured to
compute the two optimal inputs namely the expansion valve opening and the compressor speed.

22

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4182012
 The controller performances were evaluated in terms of the controller’s robustness to withstand

ed
disturbances and its ability to track the reference values. In the presence of sudden disturbances
such as the environment temperature and the heat absorbed in the evaporator, the controller
response was satisfactory since the refrigerant mass flow rates through the expansion valve and
the compressor varied accordingly as optimal values of the expansion valve opening and the

iew
compressor speed were automatically actuated to keep track of the reference settings.
The coupling effects between superheat and evaporator pressure were also considered by the MPC
since the reference values were tracked simultaneously with optimal expansion valve opening and
compressor speed whose individual control action did not tradeoff the MPC robustness to
disturbance rejection and reference tracking.

ev
Conflicts of Interest
The Authors declare no conflict of interest.
Acknowledgement

r
The Energy System Group in the Department of Electrical, Electronic and Computer Engineering

References
[1]
er
at University of Pretoria, South Africa has supported this work.

Rasmussen, B.P., Alleyne, A.G., “Dynamic modelling and advanced control of air conditioning
pe
and refrigeration systems,” Urbana, 2006, Report, University of Illinois.

[2] Stoecker, W.F., “Stability of an evaporator valve control loop,” ASHRAE Transactions, 1966, Vol.
72.
ot

[3] Najork, H., “Investigations on the dynamical behavior of evaporators with thermostatic expansion
valve,” 13th International Congress of Refrigeration, 1973, pp. 759-769.

[4] Broersen, P.M.T., Vand der Jagt, M.F.G., “Hunting of evaporators controlled by a thermostatic
tn

expansion valve,” ASME Journal of Dynamic Systems, Measurement & Control, 1980, Vol. 2, pp.
130-135.

[5] Gruhle, W.D., Isermann, R., “Modeling and control of a refrigerant evaporator,” ASME Journal of
rin

Dynamic Systems, Measurement & Control, 1985, Vol. 9, 235-240.

[6] James, K.A., James, R.W., “Transient analysis of thermostatic expansion valve for refrigeration
system evaporators using mathematical models,” Transactions of the Institute of Measurement &
Control, 1987, Vol. 9, pp. 198-205.
ep

[7] Ibrahim, G.A., “Effect of sudden changes in evaporator external parameters on a refrigeration
system with an evaporator controlled by thermostatic expansion valve,” International Journal of
Refrigeration, 2001, Vol. 24, pp. 556-576.
Pr

23

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4182012
 [8] Zhijiu, C., Ruiqi, Z., Yezheng, W., Chen, W., “Experimental investigation of a minimum stable

ed
superheat control system of an evaporator,” International Journal of Refrigeration, 2002, Vol. 25,
pp. 1137-1142.

[9] Boyd, D.L., “Capacity control of reciprocating compressors used in refrigeration systems,” Purdue
e-Pubs, 1972.

iew
[10] Tassou, S.A., Marquand, C.J., “Comparison of the performance of capacity-controlled and
conventional-controlled heat pumps,” Applied Energy, 1988, Vol. 14, pp. 241-256.

[11] Qureshi, T.Q., Tassou, S.A., “Variable speed capacity control in refrigeration systems,” Applied
Thermal Engineering, 1996, Vol. 16, pp. 103-113.

ev
[12] Yaqub, M., Zubair, S.M., “Capacity control for refrigeration and air-conditioning systems: a
comparative study,” Journal of Energy Resources Technology, 2001, Vol. 123, pp. 92-99.

r
[13] Hickman, D.A., Slupsky, J., “Development, field testing and implementation of automated,
hydraulically controlled, variable volume loading systems for reciprocating compressors,”
Proceedings of IPC ’02, 4th International Pipeline Conference, 2002, pp. 1-9.

[14] er
Huh, J.H., Brandemuehl, M.J., “Optimization of air-conditioning system operating strategies for
hot and humid climates,” Energy and Buildings, 2008, Vol. 40, pp. 1202-1213.
pe
[15] Li, D., Wu, H., Gao, J., “Experimental study on stepless capacity regulation for reciprocating
compressor based on novel rotary control valve,” International Journal of Refrigeration, 2013, Vol.
36, pp.1701-1713.

[16] Wang, Y., Jiang, Z., Zhang, J., Zhou, C., Liu, W., “Performance analysis and optimization of
ot

reciprocating compressor with stepless capacity control system under variable load conditions,”
International Journal of Refrigeration, 2018, Vol. 94, pp. 174-185.

[17]
tn

Wang, L., Liu, G.B., Zhao, Y.Y., Li, L.L., “Performance comparison of capacity control methods
for reciprocating compressors,” 9th International Conference on Compressors and their Systems,
2015, pp. 1-7.

[18] He, X.D., Liu, S., Asada, H.H., “Modeling of vapor compression cycles for multivariable feedback
rin

control of HVAC systems,” ASME Journal of Dynamic Systems, Measurement & Control, 1997,
Vol. 119, pp. 183-191.

[19] Qi, Q., Deng, S., “Multivariable control-oriented modelling of a direct expansion (DX) air
ep

conditioning (A/C) system,” International Journal of Refrigeration, 2008, Vol. 40, pp. 841-849.

[20] Qi, Q., Deng, S., “Multivariable control of air temperature and humidity in a direct expansion (DX)
air conditioning (A/C) system,” Building & Environment, 2009, Vol. 44, pp. 1659-1667.
Pr

24

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4182012
 [21] Mei, J., Zhu, B., Xia, X., “Model predictive control for optimizing indoor air temperature and

ed
humidity in a direct expansion air conditioning system,” IEEE 27th Chinese Control and Decision
Conference (CCDC), 2015, pp. 2491-2496.

[22] Mei, J., Xia, X., “Energy-efficient predictive control of indoor thermal comfort and air quality in a
direct expansion air conditioning system,” Applied Energy, 2017, Vol. 195, pp. 439-452.

iew
[23] Mei, J., “Energy Efficiency Control of Direct Expansion Air Conditioning Systems,” Pretoria,
2018, PhD Thesis, University of Pretoria.

[24] He, X.D., Liu, S., Asada, H.H., Itoh, H., “Multivariable control of vapor compression systems,”
HVAC&R Research, 1998, Vol.4, pp. 205-230.

ev
[25] Leducq, D., Guilpart, J., Trystram, G., “Low order dynamic model of a vapor compression cycle
for process control design,” Journal of Food Process Engineering, 2003, Vol. 6, pp. 67-91.

r
[26] Shah, R., Rasmussen, B.P., Alleyne, A.G., “Application of a multivariable adaptive control strategy
to automotive air conditioning systems,” International Journal of Adaptive Control and Signal
Processing, 2004, Vol. 18, pp. 199-221.

[27] er
Sanama, C., Xia, X., “Transient state modelling and experimental investigation of the thermal
behavior of a vapor compression system,” Mathematical Problems in Engineering, 2021, pp. 1-14.
pe
[28] McKinley, T.L., Alleyne, A.G., “An advanced nonlinear switched heat exchanger model for vapor
compression cycles using the moving-boundary method,” International Journal of Refrigeration,
2008, Vol. 31, pp. 1243-1264.

[29] Winkler, J.M., “Development of a component based simulation tool for the steady state and
ot

transient analysis of vapor compression systems,” Maryland, 2009, PhD Thesis, University of
Maryland.

[30]
tn

Link, R., Deschamps, C.J., “Numerical modeling of startup and shutdown transients in
reciprocating compressors,” International Journal of Refrigeration, 2011, Vol. 34, pp. 1398-1414.

[31] Pangborn, H., Alleyne, A.G., Wu, N., “A comparison between finite volume and switched moving
boundary approaches for dynamic vapor compression system modeling,” International Journal of
rin

Refrigeration, 2015, Vol. 53, pp. 101-114.

[32] Bagheri, F., “Development of efficient air conditioning and refrigeration system for service
vehicles,” Vancouver, 2016, PhD Thesis, Simon Fraser University.
ep

[33] Gnielinski, V., “New equations for heat and mass transfer in the turbulent flow in pipes and
channels,” NASA, Technical Report A, Vol. 752, 1975, pp. 22020.

[34] Dobson, M., Chato, J., “Condensation in smooth horizontal tubes,” Journal of Heat Transfer, Vol.
Pr

120, 1998, pp. 193.

25

This preprint research paper has not been peer reviewed. Electronic copy available at: https://ssrn.com/abstract=4182012