Renewable Energy 77 (2015) 64e78

Contents lists available at ScienceDirect

Renewable Energy
journal homepage: www.elsevier.com/locate/renene

Transfer functions of solar heating systems for dynamic analysis

and control design

rd Kicsiny*
Richa

n University, Pa

ter K. u. 1., 2100 Go
do
llo
}, Hungary
Department of Mathematics, Institute for Mathematics and Informatics, Szent Istva

a r t i c l e i n f o

a b s t r a c t

Article history:
Received 6 August 2014
Accepted 1 December 2014
Available online

Mathematical modelling is the theoretically established tool for developing solar heating systems, e.g.
with using transfer functions. If we know the transfer functions of the system, the outlet temperature can
be predicted as a function of the input variables (solar irradiance, inlet temperature, environment
temperatures), dynamic analysis can be carried out, and stable system control can be effectively designed
based on the well-tried methods of control engineering. For these purposes, new, validated transfer
functions for solar heating systems are worked out in this study based on a mathematical model, which
can be found in the literature and has been applied successfully in the eld. The transfer functions are
used for dynamic analysis and control design of solar heating systems. The dynamic analysis is presented
and the efciency of the proposed stable control is demonstrated with respect to a real solar heating
system.
2014 Elsevier Ltd. All rights reserved.

Keywords:
Solar heating systems
Transfer functions
Dynamic analysis
Control design
PI control

1. Introduction
Mathematical modelling is the theoretically established tool for
developing solar heating systems, e.g. with using transfer functions.
Ordinary differential equation (ODE) models are widely used as
they are relatively simple and easy to handle. Among the current
collector models, the nonlinear ODE model proposed by Perers
and Bales [1] may be the most widely used one, which is the
improved version of the quasi-dynamic model from the standard
EN 12975 [2].
If there is an external heat exchanger in the system, it can be
modelled with the well-known effectiveness-NTU method [3], or
the separate sides of the heat exchanger can be assumed to have
homogeneous temperatures and can be modelled with ODEs [4].
The pipes of the system may be modelled with ODEs (assuming
homogeneous pipe temperatures), or partial differential equations
(PDEs) (with the one-dimensional linear heat transfer PDE) [5,6].
Solar storages can be also modelled with ODEs. See Ref. [7] for
ODEs of mixed storages and stratied storages.
Georgiev [8] connected a distributed (PDE) collector model and
a mixed (ODE) storage model to describe a collector-storage

* Tel.: 36 28522000/1414; fax: 36 28410804.

E-mail address: Kicsiny.Richard@gek.szie.hu.
http://dx.doi.org/10.1016/j.renene.2014.12.001
0960-1481/ 2014 Elsevier Ltd. All rights reserved.

system. After connecting the models of the working components,

solar heating systems are generally modelled with ODEs [9e11].
In Refs. [4,12], collector-heat exchanger-storage systems are
modelled with a linear (multidimensional) ODE, which will be used
in the present paper. This model is validated [4,12,13] or partly
validated [14] and accurate enough for different successful applications [10,12,13,15] as well as its improved version in Ref. [13],
where the pipes of the system are also modelled with ODEs. The
advantage of the basic linear model of Ref. [4] is that it is simpler
and easier to use than its extended linear version [13], its nonlinear
version [16] or the delay differential equation model of Ref. [15].
The extended linear model of Ref. [13] and the model of Ref. [15] are
roughly the same precise and they are more precise than the basic
model of Ref. [4]. On the other hand, the extended linear model [13]
is simpler and easier to use than its nonlinear version [13] or the
model of [15]. Furthermore, the mentioned nonlinear models are
approximately the same precise as their linear versions (see Ref.
[13], cf [12,16] or see Ref. [17]). Thus the models of [13,15] are the
most advantageous in view of accuracy while the linear model of
[4] is the most advantageous in view of simplicity. In addition, the
latter model is the basis for all other models in Refs. [13,15,16].

s et al. [9], collector transfer functions
From the model of Buza
have been determined and applied for the dynamic analysis of real
collectors [18,19]. The present work extends these results by
determining transfer functions for whole solar heating systems and
applying them for the dynamic analysis of a real system.

R. Kicsiny / Renewable Energy 77 (2015) 64e78

Nomenclature
Ac
Ah
Ahe
cc
ci
ch
Ic
khe
mh
t

collector surface area (m2)

surface area of the heat transfer inside the heat
exchanger (m2)
surface area of the heat exchanger to the environment
(m2)
specic heat capacity of the collector uid (J/(kg K))
specic heat capacity of the uid in the inlet loop (J/
(kg K))
specic heat capacity of the heat exchanger material (J/
(kg K))
(global) solar irradiance on the plane of the collector
(W/m2)
heat loss coefcient of the heat exchanger to the
environment (W/(m2 K))
mass of the empty heat exchanger (kg)
time (s)

Generally speaking, transfer function based modelling is relatively new and rare in the analysis of solar heating systems, especially, in domestic case. Besides the latter two references, some
examples are the following: Amer et al. [20] solved a collector
model with time and one space dimension for the uid temperature using Laplace transformation. Huang and Wang [21] wrote a
nonlinear two-node collector model into Laplace transformed form
to gain transfer functions. Bettayeb et al. [22] used a two-node
model to determine collector transfer functions for the uid temperature and the absorbed solar energy.
The most prevalent and simple control strategy is the on/off
control for solar heating systems in domestic hot water (DHW)
production working with constant ow rate, see e.g. Refs. [7,23].
Several controls using pump ow modulation are used in solar
heating systems: Winn [24] compared on/off, I (integral) and PID
(proportional integral differential) controls, Hirsch [25] compared
f [26] discussed on/
on/off, P (proportional) and hybrid controls. Lo
off, P, I, PID, adaptive and certain optimal controls. Optimal controls
often maximize the overall energy gain by ow rate modulation.
See Refs. [27e29] for the case of no heat exchanger and [30] for the
case of a counter ow heat exchanger.
P and PI (proportional integral) controls for collectors are given
and a PI control is worked out in details for a real collector eld in Ref.
[19] based on proposed collector transfer functions. Based on studies
in the literature, not many improvements on control for solar heating
systems used in domestic applications have been established in the
recent few decades. In particular, transfer function based control is
rather rare. Pasamontes et al. [31] serve with a further example on
the control of the collector eld of a solar cooling system based on
the transfer function for a mathematical model with time delay.
Transfer function based control is more prevalent for industrial
processes e.g. for solar power or desalination plants [32e35].
In the present study, new, validated transfer functions for solar
heating systems used primarily for domestic purposes are proposed and used for dynamic analysis and control design. According
to a there appointed future research, the present study extends the
results of [19], where transfer functions, dynamic analysis and
controls have been proposed for solar collectors. The below worked
out transfer functions are unique concerning the linear ODE model
for solar heating systems in Ref. [4]. If the method for working out
transfer functions for this basic model is presented, corresponding

Tc
Tce
Thc
The
Thh
Ti
UL
vc
vi
Vc
Vh
kh

h0
F
rc
ri

65

collector temperature ( C)
collector environment temperature ( C)
cold side temperature in the heat exchanger ( C)
environment temperature of the heat exchanger ( C)
hot side temperature in the heat exchanger ( C)
inlet (uid) temperature of the system ( C)
overall heat loss coefcient of the collector (W/(m2 K))
(pump) ow rate in the collector loop (m3/s)
(pump) ow rate in the inlet loop (m3/s)
collector volume (m3)
volume of the heat exchanger (m3)
heat transfer coefcient inside the heat exchanger (W/
(m2 K))
collector optical efciency (-)
heat exchanger effectiveness (-)
mass density of the collector uid (kg/m3)
mass density of the uid in the inlet loop (kg/m3)

transfer functions can be derived rather straightforwardly for the

more precise extended linear model of [13] based on the same
method. That is why the linear model of [4] is essential and applied
in this study to work out transfer functions. (Because of limits in
volume, the transfer functions for the extended model of [13]
cannot be derived and detailed here.) More precisely, the truncated version of the model of [4] (without modelling the solar
storage) is used in the present study, the validation of which is also
presented below based on measured data. This means that the here
proposed transfer functions are also validated, since they form an
alternative representation of the same mathematical model.
The advantages are considerable: Knowing the transfer functions, dynamic analysis of solar heating systems can be carried out
and feedback control can be effectively designed based on the welltried methods of control engineering. A control determined in such
a way is generally much simpler than nonlinear and optimal controls and can follow the reference signal much more precisely than
the most frequently used on/off controls. Perhaps, the simple
applicability is the main advantage of the linear approach concerning transfer functions.
The contributions of the present work in details are the
following:
1. Based on a validated and successfully applied ODE model for
solar heating systems [4], new transfer functions are mathematically derived (at the end of Section 3.1) and validated. The
applicability of the transfer functions are interpreted with the
dynamic analysis of a real system.
2. As a further important application of the transfer functions,
closed-loop (PI) control design is given with stability criteria for
solar heating systems using the methods of control engineering.
The efciency of the proposed control design is demonstrated
based on simulation results.
The present paper proposes all the concepts of work [19]
(transfer functions, dynamic analysis, control design) for whole
solar heating systems and not for collectors alone, so the present
contributions are even more important and more general than the
results of [19].
The paper is organized as follows: Section 2 describes the model
for solar heating systems, for which the transfer functions are

66

R. Kicsiny / Renewable Energy 77 (2015) 64e78

determined in Section 3 and applied for the dynamic analysis of a

real solar heating system. The transfer functions are used for
designing a stable feedback control in Section 4, where the proposed control is applied and evaluated for the same real system.
Section 5 contains nal conclusions and future research proposals.
See Ref. [36] for the used concepts of control engineering (Laplace transformation, transfer function, step response, P, PI controls,
static error, stability, etc.). Maple [37], with its symbolic solver, is
used below for mathematical derivations (e.g. by means of Laplace
and inverse Laplace transformations) mainly to gain the transfer
functions. Matlab [38] is used for validation and to implement and
simulate the PI control for a real solar heating system.

This section presents the mathematical model for solar heating

systems from Ref. [4] without the equation corresponding to the
solar storage. The proposed transfer functions will be determined
based on this truncated model, the validation of which is also
carried out.
2.1. Mathematical model
Consider the solar heating system in Fig. 1.
The mathematical model of the system in Fig. 1 is the following
(see also [4]):

dTc t
Ac h0
U Ac
vc

Ic t L
Tce t  Tc t Thh t  Tc t;
dt
rc cc Vc
rc cc Vc
Vc
(1a)
dThh t
r cc vc
k A
cm c
Tc t  Thh t c m h h V Thc t
Vh
h h
h h
dt
rc cc
rc cc h
2

 Thh t c

Ahe khe =2
h mh

rc cc V2h

The t  Thh t;
(1b)

dThc t
rcv
k A
c m i i i V Ti t  Thc t c m h h V Thh t
h h
h h
dt
ri ci h
ri ci h
2

 Thc t c

Ahe khe =2
h mh

ri ci V2h

1. Here, vc is always maximal (constant) to keep the collector

temperature at an all-time minimal level. This maximizes the
efciency of the collector and the solar heating system in case of
any xed value of vi. (Otherwise, vc could be variable in the
models.) vi will be variable in Section 4, since it will be used as
manipulated variable in the proposed control.
2. Basically, the effect of wind speed is not included in the model
(1aec), nevertheless, one may consider it within the coefcient
UL (see e.g. Ref. [39]).
2.2. Validation

2. Mathematical model and validation

Remark 2.1

The t  Thc t:
(1c)

Model (1aec) is applied for a real solar heating system used for

n UniDHW production. The system is installed at the Szent Istva
do
llo
}, Hungary [40]. The system (SZIU system) has
versity (SZIU) Go
the following parameter values [12]: Ac 33.3 m2, h0 0.74,
rc 1034 kg/m3, cc 3623 J/(kg K), Vc 0.027 m3, UL 7 W/(m2 K),
ch 464.76 J/(kg K), mh 37 kg, Vh 0.005 m3, kh 2461.5 W/
(m2 K) Ah 2 m2, khe 5 W/(m2 K), Ahe 0.24 m2, ri 1000 kg/m3,
ci 4200 J/(kg K), vc 0 or 16.3 l/min (0.000272 m3/s), vi 0 or
10.5 l/min (0.000175 m3/s).
The calculations have been done in Matlab (and Matlab Simulink). Fig. 2 shows the Simulink diagram of model (1aec).
For validation, the measured values of Ti, Ic, Tce, The, vc and vi are
fed into the computer model of (1aec) as inputs as well as the initial
values of Tc, Thh and Thc then the modelled and measured values of
the outlet temperature Thc are compared. (Ti and Thc are measured on
the inlet and outlet pipe just before and after the storage side of the
heat exchanger, respectively.) Fig. 3 shows the comparison of the
modelled and measured temperatures for two days: 22nd
September 2012 and 2nd November 2012. The simultaneous operating states of the pumps (on/off) are also shown in the gure.
The average of the absolute difference between the modelled
and measured outlet temperatures is 2.6  C on 22nd September
2012 and 1.8  C on 2nd November 2012. In proportion to the difference between the maximal and minimal measured temperature
values, the time average of the absolute difference (or absolute
error) is 7.2% on 22nd September 2012 and 6.5% on 2nd November
2012 (c.f. [12]).
It can be concluded that the model tracks the physical processes
characteristically right with acceptable accuracy in view of several
engineering aims (studying and developing solar heating systems).
See e.g. Refs. [41], where the reasonability of such precision is
reinforced for similar systems. Thus we can accept and apply the
mathematical model (1aec).
Remark 2.2
In fact, the outlet temperature Thc is difcult to model because of
the small volume of the heat exchanger, which causes rather high
and fast changes in the modelled temperature mainly when the
pump ow rates are changing frequently (see Fig. 3).
3. Transfer functions
In this section, the transfer functions are derived and applied for
dynamic analysis. Here, vi is assumed to be constant.
3.1. Transfer function derivation

Fig. 1. Scheme of the solar heating system.

First, Eqs. (1aec) should be rewritten from time domain to

Laplace domain with Laplace transformation:

R. Kicsiny / Renewable Energy 77 (2015) 64e78

67

Fig. 2. Computer model of the solar heating system.


Ac h0
U Ac

sT c s  Tc 0
I c s L
T ce s  T c s
rc cc Vc
rc cc Vc

vc

T s  T c s ;
Vc hh

sT hh s  Thh 0 c

rc cc vc
h mh

rc cc V2h




T c s  T hh s

kh Ah
h mh

rc cc V2h

sT hc s  Thc 0 c

h mh

h mh

Wc3 sT ce s;

ri ci V2h

T hh s Whh0 sThh 0 Whh1 sT c s Whh2 sT hc s

Whh3 sT he s;




T hh s  T hc s


Ahe khe =2

T he s  T hc s ;
Vh
h mh

r
c
i i2
2

(3a)

(2b)




T i s  T hc s

kh Ah
2

ri ci V2h

T c s Wc0 sTc 0 Wc1 sI c s Wc2 sT hh s




T hc s  T hh s


A k =2

c m he he V T he s  T hh s ;
h h
h
2 rc cc 2
ri ci vi

(2a)

where the variables in Laplace domain are marked with overbars, s

is the independent complex variable in Laplace domain, Tc(0),
Thh(0), Thc(0) are the initial values of the state variables Tc, Thh, Thc. It
is a great advantage that the system of linear ODEs (1aec) has been
simplied to the system of linear algebraic equations (2aec).
Rearranging equations (2aec), we get:

(3b)

T hc s Whc0 sThc 0 Whc1 sT i s Whc2 sT hh s

Whc3 sT he s;
(2c)

Fig. 3. Validation of the used mathematical model.

(3c)

68

R. Kicsiny / Renewable Energy 77 (2015) 64e78

where

Wc0 s

tc
;
tc s 1

Wc1 s

Wc2 s

tc
vc
$ ;
t c s 1 Vc

Wc3 s

tc
U Ac
$ L ;
tc s 1 rc cc Vc

thh
rc c c v c
thh
kh Ah
$
$
; Whh2 s
;
Vh
V
thh s 1 ch mh
thh s 1 ch mh
rc cc
rc cc h
2
2
2
2
thh
Ahe khe =2
thc
thc
ri ci vi
$
; Whc1 s
$
; Whc0 s
;
Whh3 s
V
V
thh s 1 ch mh
thc s 1
thc s 1 ch mh
rc cc h
ri ci h
2
2
2
2
thc
kh Ah
thc
Ahe khe =2
$
$
; Whc3 s
;
Whc2 s
V
V
thc s 1 ch mh
thc s 1 ch mh
ri ci h
ri ci h
2
2
2
2
Whh0 s

thh
;
thh s 1

tc
Ac h0
$
;
tc s 1 rc cc Vc

Whh1 s

where tc, thh, thc are the so-called time constants of the collector
and the hot and cold sides of the heat exchanger, respectively:

tc

1
UL Ac
rc cc Vc

thc

Vvcc

; thh

ch mh rc cc Vh
;
2rc cc vc kh Ah Ahe khe

Thh(0), Thc(0) are assumed to be zero. Based on Eq. (4), the transfer
functions for each input are the following:

T hc s
T s
T s
T s
W1 s; hc W2 s; hc W3 s; hc W4 s:
T i s
I c s
T ce s
T he s

ch mh ri ci Vh
:
2ri ci vi kh Ah Ahe khe

The response of the output temperature to the initial conditions

can be similarly characterized with the following functions:

Solving the algebraic equations (3aec) for T hc s, we get:

T hc s Wi1 sTc 0 Wi2 sThh 0 Wi3 sThc 0 W1 sT i s

W2 sI c s W3 sT ce s W4 sT he s;
(4)
where

T hc s
T hc s
T hc s
Wi1 s;
Wi2 s;
Wi3 s:
Tc 0
Thh 0
Thc 0
According to the superposition principle of linear systems, the
resultant effect of the inputs and the initial temperatures on the
output is the simple sum of the single effects of each input and initial
temperatures. Eq. (4) expresses this principle in Laplace domain.

Wi1 s

Whc2 Whh1 Wc0

Whc2 Whh0
Whc0  1 Whh1 Wc2
; Wi2 s
; Wi3 s
;
1 Whh1 Wc2 Whc2 Whh2
1 Whh1 Wc2 Whc2 Whh2
1 Whh1 Wc2 Whc2 Whh2

W1 s

Whc1  1 Whh1 Wc2

Whc2 Whh1 Wc1
Whc2 Whh1 Wc3
; W2 s
; W3 s
;
1 Whh1 Wc2 Whc2 Whh2
1 Whh1 Wc2 Whc2 Whh2
1 Whh1 Wc2 Whc2 Whh2

W4 s

Whh3 Whc2  Whc3 Whc3 Whh1 Wc2

;
1 Whh1 Wc2 Whc2 Whh2

where, for simplicity, the independent complex variable s was not

indicated everywhere.
In systems engineering approach, the solar heating system has
an output variable Thc and input variables according to Fig. 4.
The transfer functions are the Laplace transformed form of the
output T hc s divided with the Laplace transformed form of the
inputs T i s, I c s, T ce s, T he s. When the transfer function corresponding to a given input is determined, the other inputs and Tc(0),

Fig. 4. Block diagram of the solar heating system.

Remark 3.1
The above proposed transfer functions are validated, since they
form an alternative representation of the mathematical model
(1aec) validated in Section 2.2.

3.2. Dynamic analysis

The transfer functions can be used for the dynamic analysis of
solar heating systems. The dynamic features of a system can be well
characterized with its unit step responses. The unit step response
corresponding to a given input is the response (output) of the
system to the input in time domain, supposing that the input is the
unit step input, while the other inputs and the initial values of the
state variables are zero. The unit step input is the following:

R. Kicsiny / Renewable Energy 77 (2015) 64e78


Inputt

0;
1;

t < 0;
;
t  0;

69

(5)

the Laplace transformed form of which is the following:

1
Inputs :
s

(6)

The unit step response as output can be given with the following
formula in Laplace domain using the transfer function for the corresponding input (W(s)):
Fig. 5. Output response of the SZIU system to the unit step of Ti.

1
Outputs Ws :
s

(7)

The unit step response in time domain can be determined from

Outputs:



1
Outputt L1 Ws ;
s

(8)

where L1 denotes the inverse Laplace transformation.

As a part of the dynamic analysis, the effect of the initial values
of the state variables on the system can be also investigated in time
domain based on the inverse Laplace transformed form of the
product of the given initial condition and the corresponding
transfer function (Wi(s)) (here, the inputs and the other initial
values are zero again):

Initial condition$L

Wi s:

(11)
The unit step response corresponding to Tce is the following:

Thc t 0:2  0:00019e0:58t 0:025e0:054t  0:223e0:0055t :

(12)
The unit step response corresponding to The is the following:

Thc t 0:0013 0:0000026e0:58t  0:00054e0:054t

 0:00075e0:0055t :

Outputt L1 Wi s$Initial condition

1

Thc t 0:0210:00002e0:58t 0:0026e0:054t 0:024e0:0055t :

(9)

Apply the described dynamic analysis for the solar heating

system. The unit step responses of the system corresponding to the
inputs Ti, Ic, Tce, The are the following:

The response corresponding

(Tc(0) 1  C) (see also Fig. 7):

(13)
to Tc(0)

is

the

following

Thc t 0:05e0:58t  0:577e0:054t 0:528e0:0055t :

(14)





1
1
Thc t L1 W1 s ; Thc t L1 W2 s ;
s
s




1
1
and Thc t L1 W4 s :
Thc t L1 W3 s
s
s
The responses corresponding to the initial conditions Tc(0),
Thh(0) and Thc(0) are the following:

Thc t Tc 0L1 Wi1 s; Thc t Thh 0L1 Wi2 s

and Thc t Thc 0L1 Wi3 s:

3.2.1. Application for the SZIU system

The above dynamic analysis is applied for the SZIU system. The
tap water is led into the storage side of the heat exchanger with
temperature Ti. The output uid is the required DHW with temperature Thc (see Fig. 1). Here, the pumps are considered switched
on. The response functions to each input are provided below.
Some of them are also depicted (but not all because of limits in
volume).
The unit step response corresponding to Ti is the following (see
also Fig. 5):

Fig. 6. Output response of the SZIU system to the unit step of Ic.

Thc t 0:8  0:031e0:58t  0:34e0:054t  0:43e0:0055t :

(10)
The unit step response corresponding to Ic is the following (see
also Fig. 6):

Fig. 7. Output response of the SZIU system to Tc(0) 1  C.

70

R. Kicsiny / Renewable Energy 77 (2015) 64e78

2. The above dynamic analysis is essentially theoretical, since it

requires a perfect unit step as one of the inputs while the others
are constant, furthermore, the system must be in a perfect
equilibrium at the beginning. Of course, such conditions do not
occur in practice. Nevertheless, this dynamic analysis is a
standard and widely used way for characterizing dynamic
systems.

4. System control

Fig. 8. Output response of the SZIU system to Thc(0) 1  C.

The response corresponding to Thh(0) is the following

(Thh(0) 1  C):

Thc t 0:48e0:58t 0:419e0:054t 0:065e0:0055t :

(15)

The response corresponding to Thc(0) is the following

(Thc(0) 1  C) (see also Fig. 8):

Thc t 0:466e0:58t 0:473e0:054t 0:061e0:0055t :

Stable control can be designed for solar heating systems using

the transfer functions and the methods of control engineering. The
aim is to change the outlet temperature as controlled variable according to a given reference function in time by proper modulation
of the inlet pump ow rate vi as manipulated variable. Here, vc is
constant, vi is variable (see Note 1 in Remark 2.1). In this approach,
functions Ti, Ic, Tce, The are disturbances. This control is summarized
in Fig. 9.
Now, vi(t) is not constant, so not all coefcients are constant in
system (1aec) even (1aec) is not linear in the variables Thc(t), Ti(t)
and vi(t), because of the products vi(t)Thc(t) and vi(t)Ti(t) in (1c).
Therefore, the classical linear methods of control engineering
cannot be directly applied. First of all, the model (1aec) should be
linearized in a convenient operating point.

(16)

If the different inputs and initial conditions effect at the same

time, the resultant output can be determined with a simple sum of
functions (10)e(16) according to the superposition principle:

Thc t1:0210:00002e0:58t 0:0026e0:054t 0:024e0:0055t :

(17)
For visibility, the responses in Figs. 5e8 are presented for
different time periods.
Remark 3.2
1. According to functions (10)e(16), the largest effect of the inputs on the outlet temperature Thc is evoked by the unit
change of Ti, and the lowest effect is evoked by the unit
change of The (since khe 5 W/(m2 K) is very low because of
the insulation of the heat exchanger). Among the initial conditions, Thc(0) has the largest and Thh(0) has the lowest effect.
It seems that each input has rather low effect. For example, if
Ic increases by 1 W/m2, Thc increases only by approximately
0.02  C (see Fig. 6). It is reasonable considering the rather high
ow rates vc 16.3 l/min, vi 10.5 l/min. Nevertheless, the
output is proportional to the inputs according to the linearity
principle, so higher inputs involve (proportionally) higher
output: e.g. if Ic increases by 100 W/m2, Thc increases by
approximately 2  C.

4.1. Model linearization

Practically, such an equilibrium of model (1aec) is chosen as
operating point, which represents a sort of average operating
condition, that is to say, in case of which functions Tc(t), Thh(t), Thc(t),
Ti(t), Ic(t), Tce(t), The(t) are constants, where each constant is the
approximate average between the upper and lower limits of its real
0 , T 0 , T 0 , I 0 , T 0 and T 0 denote the corpossible values. Let Tc0 , Thh
c
ce
i
hc
he
responding constants at such an operating point. The right hand
side of (1aec) are zero at the operating point, since it is an
equilibrium:

 v 

Ac h0 0
U Ac  0
c
0
Tce  Tc0
Thh
Ic L
 Tc0 ;
rc cc Vc
rc cc Vc
Vc

0 c

rc cc V_ c
h mh
2

h mh

rc cc V2h

ri ci v0i
h mh


0
Tc0  Thh
c

Ahe khe =2 

0c

rc cc V2h

ri ci V2h

ri ci V2h

rc cc V2h

0
0
Thc
 Thh


(18b)

0
0
The
;
 Thh


0
Ti0  Thc
c

Ahe khe =2 

h mh

kh Ah
h mh
2

(18a)

kh Ah
h mh

ri ci V2h

0
0
Thh
 Thc

0
0
The
;
 Thc

(18c)

The inlet ow rate at the operating point from Eq. (18c) is

v0i




.
0  T0
0  T0
Ahe khe Thc
2
kh Ah Thc
hh
he



:
0 rc
Ti0  Thc
i i

(19)

Eq. (1c) has the following form:

dThc t
f Thh t; Thc t; Ti t; The t; vi t;
dt

Fig. 9. Block diagram of the solar heating system in view of control.

(20)

based on which the linearized version of (1c) at the operating point

is the following:

R. Kicsiny / Renewable Energy 77 (2015) 64e78

71




dThc t
vf  0 0 0 0 0  
vf  0 0 0 0 0 
0
0
0
0
f Thh
Thh ; Thc ; Ti ; The ; vi $ Thh t  Thh

T ;T ;T ;T ;v $
; Thc
; Ti0 ; The
; v0i
dt
vThh
vThc hh hc i he i
 vf 
 

 vf
vf  0 0 0 0 0  
0
0
0
0
0
Thc t  Thc
; Thc
; Ti0 ; The
; v0i $ Ti t  Ti0

Thh
Thh ; Thc ; Ti ; The ; vi $ The t  The
$
vTi
vThe
vvi
. 


0
 

  ri ci vi kh Ah khe Ahe 2


kh Ah
0
0
0
0

Thh
; Thc
; Ti0 ; The
; v0i $ vi t  v0i 0 
Thh t  Thh
$
ch mh
V
ch mh
V
ri ci h
ri ci h
2
2
2
2


0
0


 ri ci Ti  Thc 



ri ci v0i
khe Ahe =2 
0
0

Ti t  Ti0
The t  The
vi t  v0i
Thc t  Thc
ch mh
V
ch mh
V
ch mh
V
ri ci h
ri ci h
ri c i h
2
2
2
2
2
2

sT~ c s
Eq. (1a,b) are linear with respect to all time-dependent functions, so the coefcients in these equations remain the same in the
linearized model below (22aec).
0 , T
~ t T t  T 0 ,
Let T~ c t Tc t  Tc0 , T~ hh t Thh t  Thh
hc
hc
hc
0
0
0,
~
~
I c t Ic t  Ic ,
T~ ce t Tce t  Tce
T i t Ti t  Ti ,
0 , ~
vi t vi t  v0i , with which the linearized
T~ he t The t  The
model is

 v 

Ac h0 ~
U Ac  ~
c ~
Ic s L
T ce s T~ c s
T hh s T~ c s ;
rc cc Vc
r c c c Vc
Vc

sT~ hh s c

(22a)

h mh

c
c

~ s
W
1

~
~
~
~
1 W
hh1 sW c2 s W hc2 sW hh2 s

(22b)

rc cc V2h

(23a)


T~ c s  T~ hh s


kh Ah
h mh
2

rc cc V2h

T~ hc s  T~ hh s


Ahe khe =2  ~
T he s  T~ hh s ;
Vh
2 rc cc 2

ri ci V2h

T~ i s  T~ hc s


kh Ah
h mh
2

ri ci V2h

T~ hh s  T~ hc s


Ahe khe =2  ~
T he s  T~ hc s
Vh
h mh
2 ri ci 2


rc
0
~vt;
c m i i V Ti0  Thc
h h
h
2 ri ci 2

(23c)

(22c)

from which the resultant effect of the inputs is the sum

of the single effects according to the linear superposition
principle:

~ sT~ s W
~ s~I c s W
~ sT~ ce s W
~ sT~ s
T~ hc s W
1
2
3
4
i
he
~ 5 s~v s;
W
i

~ s
; W
2

(24)

where

~
~
~
W
hc2 sW hh1 sW c1 s
~
~
~
~
1 W sW s W W
hh1

c2

hc2

hh2 s

~
~
~
~
~
~
~
~
~
W
hc2 sW hh1 sW c3 s
~ s W hh3 sW hc2 s  W hc3 s W hc3 sW hh1 sW c2 s;
; W
4
~
~
~
~
~
~
~
~
1 W hh1 sW c2 s W hc2 sW hh2 s
1 W hh1 sW c2 s W hc2 sW hh2 s


~
~
~
W
hc4 s  1 W hh1 sW c2 s
~ s
;
W
5
~
~
~
~
1 W
hh1 sW c2 s W hc2 sW hh2 s

~ s
W
3

(23b)

h mh

ri ci v0i
h mh

Rewrite (22aec) (with T~ c 0 T~ hh 0 T~ hc 0 0  C) into

Laplace domain:



~
~
~
W
hc1 s  1 W hh1 sW c2 s

sT~ hc s c

rc cc vc

 v 

dT~ c t Ac h0 ~
U Ac  ~
c ~

I c t L
T ce t T~ c t
T t T~ c t ;
dt
rc cc Vc
rc cc Vc
Vc hh


dT~ hh t
r cc vc
c m c
T~ c t  T~ hh t
V
h h
h
dt
2 rc cc 2


k A
c m h h V T~ hc t  T~ hh t
h h
h
2 rc cc 2

A k =2 
c m he he V T~ he t  T~ hh t ;
h h
h
2 rc cc 2
0


~
ri ci vi
dT hc t
c m
T~ i t  T~ hc t
V
h
h
h
dt
2 ri ci 2


k A
c m h h V T~ hh t  T~ hc t
h h
h
2 ri ci 2

A k =2 
c mhe he V T~ he t  T~ hc t
h h
h
2 ri ci 2


rc
0
~vi t:
c m i i V Ti0  Thc
h h
h
2 ri ci 2

(21)

72

R. Kicsiny / Renewable Energy 77 (2015) 64e78

where

~ s Wc1 s;
W
c1

~ s W s;
W
c2
c2

~ s W s;
W
c3
c3

~
W
hh1 s Whh1 s;

according to the notation in Section 3.1, and

ri ci v0i
~
thc
~
$c m
;
W
hc1 s
~
thc s 1 h h ri ci Vh
2

~
thc
kh Ah
$
;
Whc2 s
~
thc s 1 ch mh ri ci Vh
2

where

(25)

(26)

~ s
W
1
;
~
~ s
1 W c sW
5

(27)

~ s
W
2
;
~ 5 s
~
1 W c sW

(28)

~~
W
T

hc ;I c

~ s
W
3

(29)

~ s
W
4
:
~ s
~
1 W c sW
5

(30)

~ 5 s
~ c sW
1W

~ c s AP ;
W

~ c s AP 1 1 AP 1 sTI ;
W
sTI
sTI

(31)

(32)

(33)

where AP and TI are constant. Based on Section 4.1, it can be derived

~ c sW
~ 5 s is in accordance with the general form
that the product W
of Eq. (31) in case of each control type:

P:

PI :

~ c sW
~ 5 s cc;P W
~ s;
W
0
s0
~ 5 s cc;PI W
~ c sW
~ s:
W
0
s1

(34)

(35)

Let us take reference inputs of the following sort (when the

disturbances are zero):

T~ hc;r t cr t j ;

hc ;T i

~ 0 1.
where i and cc are constant and W
0
Consider the possibility of P and PI controls:

PI :

The aim is to realize a stable closed-loop control of the solar

heating system (1aec) according to Figs. 9 and 10 such that the
outlet temperature Thc(t) follows a predetermined reference input
Thc,r(t) in time precisely enough. This requirement corresponds to
0 .
that T~ hc t follows T~ hc;r t, where T~ hc;r t Thc;r t  Thc
~ c such that the
In effect, the problem is the determination of W
control is stable with conveniently small static errors corresponding to the inputs T~ hc;r , T~ i , ~I c , T~ ce , T~ he . The transfer functions of the
control in Fig. 10 with respect to the reference input and the disturbances T~ i , ~I c , T~ ce , T~ he are the following:

~~
W
T

hc ;T he

P:

4.2. Control design

~
~
~ ~ ~ s W c sW 5 s ;
W
T hc ;T hc;r
~ c sW
~ 5 s
1W

~~
W
T

~ 5 s cc W
~ c sW
~ s;
W
0
si

ch mh ri ci Vh

~
thc 
:
2 ri ci v0i kh Ah Ahe khe

hc ;T ce

The so-called loop gain of the controlled system (the gain in

~ c sW
~ 5 s. Let us write it
going around the feedback loop) is W
below (in (34) and (35)) in the general transfer function form
cc ~
W 0 s:
si

~
thc
A k =2
$ he he
;
Whc3 s
~
thc s 1 ch mh ri ci Vh
2
2


0  T0
T
r
c
i
i
~
thc
i
hc
$
;
Whc4 s
~
thc s 1 ch mh ri ci Vh
2

~~
W
T

~
~
W
hh2 s Whh2 s; W hh3 s Whh3 s;

(36)

where j and cr are constant. If j 0, T~ hc;r t is step function, if j 1, it

is ramp function (only t  0 is of importance in our case).
If i > j holds for i and j in (31) and (36), the static error of the
control corresponding to T~ hc;r is zero. The control is required to
follow the step inputs (j 0) precisely, that is with zero static error,
so the PI control is chosen, since i 1 > j 0 (see (35)) is fullled in
this case. Thus cc cc,PI, where cc,PI can be found in the Appendix as
Eq. (A1) (which can be derived based on Section 4.1).
The static error of the PI control in case of ramp function as
reference input (j 1) is the following:


 c
r
er;s lim T~ hc;r t  T~ hc t :
t/
cc

(37)

Consider the disturbance T~ i t in the following form (when

~
T hc;r t and the other disturbances are zero):

T~ i t c1 t k ;
Fig. 10. Closed-loop control of the solar heating system.

(38)

R. Kicsiny / Renewable Energy 77 (2015) 64e78

where k and c1 are constant. If k 0, T~ i t is step function, if k 1 it

is ramp function.
~ s corresponding to T~ should be
The transfer function W
1
i
considered in the following form:

~ s
W
1

cT~
sl

~ T i s;
W
0

(39)
~

~ T i s 1.
where cT~ i is constant and W
0
If i > k l is fullled for i, k and l in (31), (38) and (39), the static
error of the control corresponding to T~ i is zero.
~ s is in accordance with (39),
It can be derived similarly that W
1
where l 0 and cT~ i is according to (A2) and (A3) in the Appendix.
Thus the static error of the control is zero if T~ i t is step function
(k 0). If T~ i t is ramp function (k 1), the static error corresponding to T~ i t is the following:


 c~
T
e1;s lim T~ hc;r t  T~ hc t i c1 :
t/
cc

(40)

Consider ~I c t, T~ ce t and T~ he t similarly as in (38):

~I c t c t m ; T~ ce t c t n ; T~ t c t q ;
2
3
4
he
where m, n, q, c2, c3 and c4 are constant.
It can be shown like above that the static error corresponding to
~I c (e2,s), T~ ce (e3,s) or T~ (e4,s) are zero if m 0, n 0, q 0,
he
respectively. If ~I c t, T~ ce t or T~ he t is ramp function (m 1, n 1 or
q 1, respectively), the corresponding static errors are the
following, respectively:

e2;s
e3;s

e4;s

c~Ic

c ;
ccT~cce 2
c ;
cc 3
cT~

he

cc

(41)
(42)

c4 ;

(43)

where c~Ic , cT~ ce and cT~ are according to (A3)e(A6) in the Appendix.
he
The control parameters AP and TI should be such that the absolute values of the above static errors are not higher than a prexed positive limit E, based on the constants cr, cT~ i , c~Ic , cT~ ce , cT~ , c1,
he
c2, c3, c4 and that cc cc,PI.
Furthermore, AP and TI should be such that the control is stable.
~~ ~
Consider the transfer function with respect to T~ hc;r that is W
T hc ;T hc;r
(see (26)). The controlled system is stable corresponding to T~ hc;r if
~~ ~
the real parts of the zeros of the denominator of W
are
T hc ;T hc;r
~ c sW
~ 5 s
negative. It can be derived that the denominator 1 W
has the following form:

dr4 s4 dr3 s3 dr2 s2 dr1 s1 dr0 :

(44)

The constants dr4, dr3, dr2, dr1, dr0 can be found in the Appendix
((A7)e(A11)).
The following system of conditions is sufcient for the stability
according to the Routh-Hurwitz criterion:

dr4 > 0; dr3 > 0; dr2 > 0;


 dr3 dr1



 dr3 dr1 



 dr4 dr2  > 0;  dr4 dr2
 0 dr3

dr1 > 0;

0 
dr0  > 0;
dr1 

73

In sum, the mathematical task of designing a PI control is the

following:
  determine

 the free control
 parameters


 AP and TI such
that er;s   E, e1;s   E, e2;s   E, e3;s   E, e4;s   E hold, and
conditions (SC) hold.
Remark 4.1
The Routh-Hurwitz criterion assures the stability of not only the
linearized but the corresponding nonlinear controlled system
(where vi(t) is not constant), since the latter is also stable if the real
~ ~ ~ are negative acparts of the zeros of the denominator of W
T hc ;T hc;r
cording to Lyapunov.
4.2.1. Application for the SZIU system
Let us give an appropriate PI control for the SZIU system.
0 55  C, which is
Consider an end-of-spring period in May. Let Thc
0 56  C (a bit
high enough for general domestic purposes. Let Thh
0 ), T 0 15  C (approximate average tap water temhigher than Thc
i
0 20  C (approximate average daytime environment
perature), Tce
0 25  C (approximate average temperature in the
temperature), The
maintenance house). From these data, the remaining equilibrium
values Tc0 , Ic0 and v0i can be determined according to (18aec):
Tc0 60.85  C, Ic0 586.9 W/m2 (which is in accordance with the
approximate average daytime solar irradiance of a clear day in May
in Hungary [42]), v0i 0.000032 m3/s (1.92 l/min), the maximal
value of vi is 10.5 l/min (see Section 2.2). As further limitation, it is
assumed that vi(t) can vary between zero and the maximal value in
3 s that is the absolute value of the speed of ow rate changing (the
absolute value of the acceleration of the ow) is 0.000058 m3/s2.
Let the absolute values of the static errors (37), (40)e(43) be less
or equal than 0.1  C, which is convenient for a DHW heating
application. Let the controlled system be stable that is conditions
(SC) are satised.
Suppose that such fast changing disturbances (with high amplitudes) act on the controlled system simultaneously, which are
rare even separately in real circumstances, thus they are even more
improbable in the same time. Let us check in such a case whether
the controlled system is still able to follow precisely enough a
relatively fast changing reference input (with high amplitude). If
the controlled system can follow well such an extreme reference
input under such extreme disturbances (which have small probabilities), it will work even more satisfactorily under real circumstances. Thus the practical applicability of the control will be
thoroughly supported.
Sinusoidal inputs are given as the mentioned reference input
and disturbances (see Figs. 11 and 12) with initial values corresponding to the above operating point of the SZIU system:
0 55
Thc;r 0 T~ hc;r 0 Thc
gradient 0.5  C/min;

 C,

amplitude 5

 C,

d
 r0 > 0;

 dr3 dr1 0 0 


 dr4 dr2 dr0 0 


 0 dr3 dr1 0  > 0:


 0 d

r4 dr2 dr0
(SC)

~ c sW
~ 5 s, so (SC)
The denominators in (27)e(30) are also 1 W
is sufcient for the stability of the controlled system corresponding
to T~ i , ~I c , T~ ce and T~ he as well.

Fig. 11. Reference input to the controlled solar heating system.

initial

74

R. Kicsiny / Renewable Energy 77 (2015) 64e78

Fig. 12. Disturbances to the controlled solar heating system.

Ti 0 T~ i 0 Ti0 15  C, amplitude 5  C,
gradient 3  C/min;
Ic 0 ~I c 0 Ic0 586.9 W/m2, amplitude 150 W/m2,
gradient 20 W/(m2s);
0 20  C, amplitude 8  C,
Tce 0 T~ ce 0 Tce
gradient 1  C/min;
0
The 0 T~ he 0 The
25  C, amplitude 3  C,
gradient 0.5  C/min.

initial
initial
initial
initial

The control parameters are xed AP 0.0000086, TI  0.4,


er;s  
with which
the stability
is assured,
and
the requirements








0.1  C, e1;s   0.1  C, e2;s   0.1  C, e3;s   0.1  C, e4;s   0.1  C hold

even in case of maximal speed of changing of the inputs (when

cr 0.5  C/min, c1 3  C/min, c2 20 W/(m2 s), c3 1  C/min,
c4 0.5  C/min).
Let us apply and test the above PI control for the original e not
linearized e model (1aec) in case of the SZIU system. The model
and the control have been fed into computer in Matlab (and Matlab
Simulink) [38] environment according to Fig. 13.
The initial state variables are set Tc(0) Thc(0) Thh(0) 20  C,
hence the control has to reconcile relatively high initial error:
Thc,r(0)  Thc(0) 35  C. Figs. 14 and 15 show the simulation results.
Fig. 14 shows the reference input temperature Thc,r(t), the controlled
outlet temperature Thc(t) and the manipulated pump ow rate vi(t).

Fig. 13. Computer model of the controlled solar heating system.

R. Kicsiny / Renewable Energy 77 (2015) 64e78

75

Fig. 14. Thc,r(t), Thc(t) (controlled variable) and vi(t) (manipulated variable).

Fig. 15 shows the control error Thc,r(t)  Thc(t).

The results show that the absolute value of the control error
decreases permanently below 5% of the initial error (below
1.75  C) within 25.0 min, the absolute value of the error decreases
permanently below the prescribed limit 0.1  C within 26.4 min
(settling time). Based on these values, the worked out PI control is
considered satisfactorily fast and accurate in view of the control
purpose (even in case of extreme reference input and
disturbances).
Remark 4.2
1. On an average clear day in May, at least 440 W/m2 solar irradiance is still expected for 5e5.5 h after the above settling time
26.4 min [41]. 436.9 W/m2 is the minimal irradiance in the
simulation (see Fig. 12), which proved to be still enough to
maintain the outlet temperature at the desired level (55  C)
according to simulation experiences. Thus the controlled outlet

temperature Thc(t) should be able to follow the reference input

(precisely enough) for at least 5e5.5 h.
2. Based on simulation data, 591.5 L DHW, at an average temperature of 55  C, has been produced in 5.5 h after the above
settling time 26.4 min (on an average clear day in May),
which covers 30% of the daily demand (1990 l [6]) of the
consumer of the SZIU system (a kindergarten). The produced
DHW can be stored in a water storage, which can be discharged during the day according to the all-time hot water
demand.
3. The gained simulation results, where the control error converges to zero, underlie the statement of Remark 4.1 corresponding to the stability of the nonlinear controlled system
(where vi(t) is not constant), since the nonlinear system model
was controlled above (see Fig. 13).
4. The control error converges to zero with oscillation (see Fig. 15),
which is normal in case of PI controllers, since the manipulated
variable vi(t) is partially proportional to the time integral of the
error (not only directly to the error). This causes recurring
overruns/oscillation. In fact, vi(t) cannot be proportional to (the
integral of) the error every time because of its lower (0 l/min)
and upper (10.5 l/min) limits (see Fig. 14). That is why the graph
of the control error is not smooth but has breaking points (see
Fig. 15).
5. Conclusion

Fig. 15. Control error Thc,r(t)  Thc(t).

Generally speaking, transfer function based modelling is relatively new and rare in the analysis of solar heating systems, especially, in domestic case. In addition, not many improvements on
control for solar heating systems used in domestic applications
have been established in the recent few decades. In particular,
transfer function based control design is rather rare despite of the
simple applicability, which is the main advantage of the linear
approach concerning transfer functions. Controls based on transfer
functions are generally much simpler than nonlinear and optimal
controls and can follow the reference signal much more precisely
than the most frequently used on/off controls. This paper intended
to contribute to full these research gaps by proposing new,

76

R. Kicsiny / Renewable Energy 77 (2015) 64e78

validated transfer functions based on a validated system model and

designing stable controls (in particular, a closed-loop PI control) by
means of the transfer functions.
The transfer functions have been also applied for the dynamic
analysis of a real solar heating system, for which the proposed
stable PI control has been applied to make the outlet temperature
follow a reference input.
According to a there appointed future research, this study has
extended the results of [19], where transfer functions, dynamic
analysis and controls have been proposed for solar collectors. Since
the present paper proposes all of these concepts for whole solar
heating systems, its contributions are even more important and
more general than the results of [19].
As the main ndings of the present paper, it can be stated that
the proposed transfer functions are successful and relatively easy to
apply (see the applications in the paper) for dynamic analysis and
stable control design. Furthermore, the proposed PI control is
appropriate for the corresponding control purpose because of its
precision and rapidity even in case of fast changing reference input
and disturbances with high amplitudes.

points, and can be also used for dynamic analysis and control
design.
Acknowledgement
The author thanks the Editor for the encouraging help in the
submission process, the anonymous Referees for their valuable

n Farkas and the Department of Physics
comments and Prof. Istva
and Process Control (SZIU) for the measured data. The author
also thanks his colleagues in the Department of Mathematics in
the Faculty of Mechanical Engineering (SZIU) for their
contribution.
Appendix
The equations below have been derived based on the relations
in the main text (with the symbolic solver of Maple) and are
referred to in the text, where needed. For example, cc,PI has been
~ 5 s and W
~ c s in Sections 4.1
derived based on the expressions of W
and 4.2.



.

0
Ahe khe cc rc vc 2kh Ah UL Ac 2kh Ah rc cc vc 2cc rc vc UL Ac Ahe khe UL Ac
cc;PI  2AP ri ci  Ti0 Thc
 
TI 4rc cc vc UL Ac kh Ah 2rc cc vc UL Ac khe Ahe 4rc cc vc UL Ac ci ri v0i 4kh Ah rc cc vc khe Ahe 2ri ci $

(A1)

UL Ac khe Ahe v0i A2he k2he UL Ac 2khe Ahe rc cc vc ri ci v0i A2he k2he rc cc vc 4kh Ah rc cc vc ri ci v0i 4kh $

:
UL Ac khe Ahe 4kh Ah UL Ac ri ci v0i

In essence, the introduced dynamic analysis can be easily

adapted for any solar heating system with an external heat
exchanger. The proposed control design can be also comprehensively applied for solar heating applications, where the output
temperature should follow some reference signal in time (e.g. solar

den

cT~ i 2ri ci v0i Ahe khe 2kh Ah 2UL Ac cc rc vc

Ahe khe 2kh Ah UL Ac =den;

(A2)

where






2Ahe khe v0i 4UL Ac v0i 4kh Ah v0i cc rc vc 2Ahe khe v0i 4kh Ah v0i UL Ac ci ri 4kh Ah




2Ahe khe Ac UL A2he k2he 4kh Ah Ahe khe cc rc vc A2he k2he 4kh Ah Ahe khe Ac UL :



power and desalination plants).

The manipulated variable was the ow rate in the inlet loop. The
inlet temperature may be also used as manipulated variable,
although, this solution is very rare. On the contrary, variable ow
rate pumps are comprehensively applied in the practice of solar
heating, and more generally, in building systems.
Further research work may deal with the derivation of transfer
functions corresponding to the more precise extended linear model
of [13]. Such a work can follow in essence the steps of the derivation
for the essential basic model of [4] presented above. Furthermore,
so-called describing functions may be determined corresponding
to nonlinear mathematical models of solar heating systems.
Describing functions can be gained by means of harmonic linearization, which is a linearization method (differing from the one
applied in this paper) corresponding to convenient operating

(A3)

c~Ic 4vc kh Ah h0 Ac rc cc =den;

(A4)

cT~ ce 4vc kh Ah UL Ac rc cc =den;

(A5)

cT~ Ahe khe 4kh Ah 2UL Ac cc rc vc

he

Ahe khe 4kh Ah UL Ac Ahe khe =den:

dr4 TI Vc cc ch mh ri ci Vh rc ch mh rc cc Vh ;

(A6)
(A7)

R. Kicsiny / Renewable Energy 77 (2015) 64e78

dr3

dr2

dr1





0
0 2 2
2ch mh rc cc Vc 2r2c c2c Vh Vc Ti0  2Thc
ch mh rc cc Vc  2Thc
rc cc Vh Vc ci ri AP
2Vc v0i vc Vh $







rc cc UL Ac Vh ch mh 2vc 2v0i Vh Vc Vh2 vc c2c r2c Ahe khe 2kh Ah Vh Vc Vh2 UL Ac rc cc ci $






ri rc cc vc UL Ac c2h m2h vc Vh 2vc Vc c2c r2c 2Ahe khe 4kh Ah Vc UL Ac Vh rc cc ch mh



Ahe khe 2kh Ah Vh Vc c2c r2c TI ;

77



(A8)








2UL Ac 2rc cc vc ch mh 2vc Vh 4vc Vc c2c r2c 2Ahe khe 4kh Ah Vc 2UL Ac Vh rc cc $






0
0
0
0
0
mh ch  4Thc
Ti0  2rc cc vc Thc
 2UL Ac Thc
vc Vc  2Thc
vc Vh c2c r2c
 4kh Ah  2Ahe khe Thc
Vc






0
2Thc
Vh UL Ac rc cc ri ci AP 2rc cc vc v0i 2UL Ac v0i ch mh 2Vh vc v0i 4vc Vc v0i c2c r2c 4kh Ah v0i


 




2Ahe khe v0i Vc 2vc 2v0i Ac UL 2kh Ah vc Ahe khe vc Vh cc rc Ahe khe 2kh Ah Ac UL Vh ci ri





4kh Ah vc 2Ahe khe vc 2vc UL Ac cc rc 4kh Ah 2Ahe khe Ac UL mh ch 4kh Ah vc 2Ahe khe vc Vc
 







2kh Ah vc 2Ahe khe vc Vh c2c r2c A2he k2he 4kh Ah Ahe khe Vc Ahe khe 2kh Ah Ac UL Vh cc rc TI



0
0 2 2
2ch mh rc cc Vc 2c2c r2c Vh Vc Ti0  2Thc
ch mh rc cc Vc  2Thc
cc rc Vh Vc ci ri AP ;

(A9)






2Ahe khe 4kh Ah 4UL Ac vc cc rc 2Ahe khe 4kh Ah UL Ac Ti0 4UL Ac  2Ahe khe






0
0
ci ri AP 4Ac UL v0i 4kh Ah v0i 2Ahe khe v0i $
4kh Ah Thc
vc cc rc 4kh Ah  2Ahe khe Ac UL Thc






vc cc rc 2Ahe khe v0i 4kh Ah v0i Ac UL ci ri 2Ahe khe 4kh Ah Ac UL A2he k2he 4kh Ah Ahe khe vc $








2rc cc vc 2Ac UL mh ch 4Vc 2Vh vc c2c r2c 2Ahe $
cc rc A2he k2he 4kh Ah Ahe khe Ac UL TI







0
0
0
0
UL Ac  2Thc
rc cc vc mh ch  2Thc
Vh  4Thc
Vc vc c2c r2c $
khe 4kh Ah Vc 2UL Ac Vh rc cc Ti0  2Thc




0
0
Vc  2Thc
Vh UL Ac rc cc ci ri AP ;
 4kh Ah  2Ahe khe Thc

(A10)







0
Ahe khe cc rc vc 2kh Ah UL Ac
dr0 2AP ri ci  Ti0 Thc

2kh Ah rc cc vc 2cc rc vc UL Ac Ahe khe UL Ac :
(A11)
References
[1] Perers B, Bales C. A solar collector model for TRNSYS simulation and system
testing. A report of IEA SHC Task 26 Solar Combisystems. 2002.
[2] CEN. European Standard Draft. European Committee for Standardisation;
1997. prEN 12975e2.
[3] Kays WM, London AL. Compact heat exchangers. New York: McGraw-Hill;
1984.

s J, Farkas I. Solar domestic hot water system simulation using block[4] Buza
oriented software. In: The 3rd ISES-Europe solar World Congress (Eurosun
2000), Copenhagen, Denmark, CD-ROM Proceedings; 2000. p. 9.
[5] Orbach A, Rorres C, Fischl R. Optimal control of a solar collector loop using a
distributed-lumped model. Automatica 1981;17(3):535e9.
[6] Kicsiny R, Farkas I. Improved differential control for solar heating systems. Sol
Energy 2012;86:3489e98.
[7] Dufe JA, Beckman WA. Solar engineering of thermal processes. New York:
John Wiley and Sons; 2006.

[8] Georgiev A. Simulation and experimental results of a vacuum solar collector

system with storage. Energy Convers Manag 2005;46:1423e42.

A, Ne

meth R. Modelling and simulation of a solar
[9] Buz

as J, Farkas I, Biro
thermal system. Math Comput Simul 1998;48:33e46.
[10] Kumar R, Rosen MA. Thermal performance of integrated collector storage
solar water heater with corrugated absorber surface. Appl Therm Eng
2010;30:1764e8.
[11] Zeghib I, Chaker A. Simulation of a solar domestic water heating system.
Energy Procedia 2011;6:292e301.
[12] Kicsiny R, Varga Z. Real-time state observer design for solar thermal heating
systems. Appl Math Comput 2012;218:11558e69.

ki Cs. Extended ordinary differential equation models
[13] Kicsiny R, Nagy J, Szalo
for solar heating systems with pipes. Appl Energy 2014;129:166e76.
[14] Kicsiny R. Multiple linear regression based model for solar collectors. Sol
Energy 2014;110:496e506.
[15] Kicsiny R. New delay differential equation models for heating systems with
pipes. Int J Heat Mass Transf 2014;79:807e15.
[16] Kicsiny R, Varga Z. Real-time nonlinear global state observer design for solar
heating systems. Nonlinear Anal Real World Appl 2013;14:1247e64.
[17] Kicsiny R. Real-time state observers for solar heating systems. Mech Eng Lett
2013;10:60e6.
[18] Buz

as J. Block-oriented modeling of solar thermal systems. Thesis of the
do
llo
}: Szent Istva

n University; 2009. p. 33
doctoral (Ph.D.) dissertation. Go
[04.11.2014], http://szie.hu//le/tti/archivum/Buzas_thezis.pdf.
[19] Buz

as J, Kicsiny R. Transfer functions of solar collectors for dynamical analysis
and control design. Renew Energy 2014;68:146e55.

78

R. Kicsiny / Renewable Energy 77 (2015) 64e78

[20] Amer EH, Nayak JK, Sharma GK. A new dynamic method for testing solar at-plate
collectors under variable weather. Energy Convers Manag 1999;40:803e23.
[21] Huang BJ, Wang SB. Identication of solar collector dynamics using physical
model-based approach. J Dyn Syst Meas Control 1994;116:755e63.
[22] Bettayeb M, Nabag M, Al-Radhawi MA. Reduced order models for at-plate
solar collectors. In: GCC Conference and Exhibition (GCC), 2011 IEEE; 2011.
p. 369e72.
[23] Kalogirou SA. Solar energy engineering. Academic Press; 2009.
[24] Winn CB. Controls in solar energy systems. Advances in solar energy. American Solar Energy Society Inc.; 1983.
[25] Hirsch UT. Control strategies for solar water heating systems. Master of Science Thesis. University of Wisconsin-Madison; 1985.
f G. Active solar systems. MIT Press; 1993.
[26] Lo
[27] Kovarik M, Lesse PF. Optimal control of ow in low temperature solar heat
collectors. Sol Energy 1976;18:431e5.
[28] Winn CB, Hull DE. Optimal controllers of the second kind. Sol Energy 1979;23:
529e34.
[29] Badescu V. Optimal control of ow in solar collector systems with fully mixed
water storage tanks. Energy Convers Manag 2008;49:169e84.
[30] Hollands KGT, Brunger AP. Optimum ow rates in solar water heating systems
with a counterow exchanger. Sol Energy 1992;48:15e9.


[31] Pasamontes M, Alvarez
JD, Guzm

an JL, Lemos JM, Berenguel M. A switching control
strategy applied to a solar collector eld. Control Eng Pract 2011;19:135e45.
[32] Ayala CO, Roca L, Guzman JL, Normey-Rico JE, Berenguel M, Yebra L. Local
model predictive controller in a solar desalination plant collector eld. Renew
Energy 2011;36:3001e12.

[33] Roca L, Guzman JL, Normey-Rico JE, Berenguel M, Yebra L. Robust constrained
predictive feedback linearization controller in a solar desalination plant collector eld. Control Eng Pract 2009;17:1076e88.
[34] Roca L, Guzman JL, Normey-Rico JE, Berenguel M, Yebra L. Filtered Smith
predictor with feedback linearization and constraints handling applied to a
solar collector eld. Sol Energy 2011;85:1056e67.
[35] Fontalvo A, Garcia J, Sanjuan M, Padilla RV. Automatic control strategies for
hybrid solar-fossil fuel plants. Renew Energy 2014;62:424e31.
[36] Bakshi UA, Bakshi VU. Linear control systems. India: Technical Publications
Pune; 2007.
[37] Maplesoft. Maple 9 learning guide. Waterloo Maple Inc.; 2003.
[38] Etter DM, Kuncicky D, Moore H. Introduction to MATLAB 7. Springer; 2004.
[39] Amrizal N, Chemisana D, Rosell JI, Barrau J. A dynamic model based on the
piston ow concept for the thermal characterization of solar collectors. Appl
Energy 2012;94:244e50.

s J, La

gym


r I, Kaboldy E, Nagy L. In: Frankovic B,
[40] Farkas I, Buza
anyosi A, Kalma
editor. A combined solar hot water system for the use of swimming pool and
kindergarten operation, energy and the environment, vol. I. Opatija: Croatian
Solar Energy Association; 2000. p. 81e8.
[41] Kalogirou SA. Applications of articial neural-networks for energy systems.
Appl Energy 2000;67:17e35.
[42] http://energetika.13s.hu/pub/_epuletenergetika_szakirany_/megujulo%
20energiaforrasok/Naplopo/EPGEP_Napkollektorok-1%5B1%5D.ppt [in Hungarian, 04.11.2014].