Applied Energy: Chang-Soon Kang, Chang-Ho Hyun, Mignon Park

Download as pdf or txt
Download as pdf or txt
You are on page 1of 14

Applied Energy 155 (2015) 270–283

Contents lists available at ScienceDirect

Applied Energy
journal homepage: www.elsevier.com/locate/apenergy

Fuzzy logic-based advanced on–off control for thermal comfort


in residential buildings
Chang-Soon Kang a, Chang-Ho Hyun b, Mignon Park a,⇑
a
Dept. of Electrical & Electronic Engineering, University of Yonsei, Republic of Korea
b
School of Electrical Electronic and Control Engineering, Kongju National University, Republic of Korea

h i g h l i g h t s

 Fuzzy logic-based advanced on–off control is proposed.


 An anticipative control mechanism is implemented by using fuzzy theory.
 Novel thermal analysis program including solar irradiation as a factor is developed.
 The proposed controller solves over-heating and under-heating thermal problems.
 Solar energy compensation method is applied to compensate for the solar energy.

a r t i c l e i n f o a b s t r a c t

Article history: In this paper, an advanced on–off control method based on fuzzy logic is proposed for maintaining ther-
Received 14 November 2013 mal comfort in residential buildings. Due to the time-lag of the control systems and the late building
Received in revised form 23 May 2015 thermal response, an anticipative control mechanism is required to reduce energy loss and thermal dis-
Accepted 25 May 2015
comfort. The proposed controller is implemented based on an on–off controller combined with a fuzzy
algorithm. On–off control was chosen over other conventional control methods because of its structural
simplicity. However, because conventional on–off control has a fixed operating range and a limited ability
Keywords:
for improvements in control performance, fuzzy theory can be applied to overcome these limitations.
Fuzzy control
Solar energy
Furthermore, a fuzzy-based solar energy compensation algorithm can be applied to the proposed con-
Energy saving troller to compensate for the energy gained from solar radiation according to the time of day.
Hydronic radiant floor heating systems Simulations were conducted to compare the proposed controller with a conventional on–off controller
under identical external conditions such as outdoor temperature and solar energy; these simulations
were carried out by using a previously reported thermal analysis program that was modified to consider
such external conditions. In addition, experiments were conducted in a residential building called Green
Home Plus, in which hydronic radiant floor heating is used; in these experiments, the proposed system
performed better than a system employing conventional on–off control methods.
Ó 2015 Elsevier Ltd. All rights reserved.

1. Introduction clean ducts, and noiseless operation [1]. In addition, RFH systems
are not only more energy-efficient than CH systems, but also pro-
Radiant floor heating (RFH) systems have gradually increased in vide more comfortable indoor thermal environments [2].
popularity because they have many advantages compared to con- Despite the many advantages of RFH systems, their
vective heating (CH) systems. The advantages of RFH systems over thermal control systems tend to suffer from problems regarding
CH systems are their low system temperatures, uniform tempera- late thermal response and management of thermal loads in
ture distributions, efficient use of space, elimination of the need to well-insulated buildings [1]. First, late thermal response arises
mainly from the thermal mass and structure characteristics of
RFH systems. Different materials have different heat transfer char-
⇑ Corresponding author at: B723, 2nd Engineering Building, Yonsei Univ.,
acteristics such as thermal mass. Concrete, is widely used material
Sinchon-dong, Seodaemun-gu, Seoul, Republic of Korea. Tel.: +82 02 2123 2868;
fax: +82 312 2333. for building construction, has a high thermal mass, increasing the
E-mail address: [email protected] (M. Park). delay between heat supply and indoor temperature response,

http://dx.doi.org/10.1016/j.apenergy.2015.05.119
0306-2619/Ó 2015 Elsevier Ltd. All rights reserved.
C.-S. Kang et al. / Applied Energy 155 (2015) 270–283 271

and thereby resulting in overheating and additional heat loss. RHF network (ANN) models for anticipative control, which determines
systems also include components such as boilers, pumps, and pipe the proper on–off times for the circulation pump in a floor heating
systems, which also introduce delay in the thermal response. circuit that supplies heat to apartment rooms. However, they con-
The other main thermal control problem of RHF systems occurs ducted this study using a simple test chamber, meaning that exter-
in well-insulated buildings; in this case, room temperature can nal factors such as sunlight could not be considered. Argiriou et al.
vary greatly depending on changes in internal loads such as the [11,12] developed an ANN-based controller for application to RFH
number of occupants or appliances in use, the type and amount systems; it predicts outdoor temperature, solar radiation, indoor
of lighting fixtures, and external loads such as solar irradiation. temperature, and supply temperature. However, due to the com-
In particular, the room temperature depends on solar irradiation plex and multi module structure of ANN, the system has a high
more than any other factor. Solar irradiation varies throughout computational cost and lacks a knowledge representation, mean-
the day of a day and can contribute a great deal of radiant energy, ing that the knowledge of expert control designers cannot be incor-
and thus it can result in overheating and thermal discomfort unless porated in its controller’s design. Furthermore, the explanation
it is appropriately taken into consideration as an important factor ability of ANN is low, meaning that the control designers cannot
in system control. sufficiently explain why they designed an ANN in a particular
There have been many reports on the topic of thermal control of way. Although some studies included the use of anticipative
RFH systems, and both on–off and proportional-integral-derivative control to enhance the control performance of an RFH system com-
(PID) control methods have been widely implemented because pared with a conventional control, many issues remain to be
they have simple structures [3–5]. However, conventional control addressed, including the problem of solar gain and optimization
methods do not sufficiently address the problem of late thermal of heating time.
response in RFH systems. Although a control system gives appro- To the best of our knowledge, no fuzzy-based thermal control
priate inputs to the building immediately depending on current method has been proposed to determine proper on–off control
condition, the building’s thermal condition does not change imme- for RFH systems, even though fuzzy logic does not require a com-
diately due to the characteristics of its thermal mass. Besides, con- plex learning process and allows the incorporation of expert
trol systems have time lag. Even if the controller is activated knowledge, unlike ANN. Some studies have investigated the effects
instantly, it takes some time to generate appropriate inputs. of solar radiation, but none have used a fuzzy-based solar energy
To solve these problems, an anticipative algorithm has been compensation algorithm at the controller level by means of chang-
researched [6–8]. Lee et al. [9,10] developed artificial neural ing the weight of the output membership function. In addition,

Fig. 1. Schematic diagram of water flow in hydronic radiant floor heating system.

Fig. 2. Schematic diagram of data flow in a hydronic radiant floor heating system.
272 C.-S. Kang et al. / Applied Energy 155 (2015) 270–283

many questions regarding practical usage remain open because two-position control is used to actuate on–off valves to maintain
most designs proposed in previous studies have been verified only the temperature within a certain deviation from set points. In the
through simulations. Other studies included experiments in test latter, the water flux is continuously controlled by means of flow
chambers or in hardware-in-the-loop-system, but they did not control valves, which are regulated by a control signal generated
cover actual disturbances in RFH systems, such as solar irradiation. based on a linear relationship between flux and heat. On/off con-
Ahn and Song [13] proposed a thermal analysis program for use trol is widely used in residential buildings because of its hardware
as a simulation tool to test the controllers for radiant floor heating simplicity; however there is a limited ability to improve the perfor-
systems. This program is very useful to verify various controllers’ mance of this control method. Thus, in the present work, an on/off
performance under consistent conditions of outdoor temperature, control was combined with a fuzzy-based control to take simulta-
solar irradiation, building materials, etc.; however, it does not neous advantage of both types of controllers.
include the solar energy model that is the main issue under consid-
eration in this paper. Hence, in the present work the thermal anal-
2.2. Test site
ysis program of Ahn was modified to include the solar energy
model presented herein.
The test facility was multi-dwelling building called Green Home
The goal of the present study was to develop a fuzzy-based con-
Plus located on the Yonsei University Global Campus in Incheon,
troller including the effects of direct sunlight, which would thereby
Korea. This facility consists of 4 floors of 5 units each; each unit
be able to solve the problem of thermal mass and thermal loads in
has 4 rooms. Fig. 3 shows photos of the building including an out-
well-insulated buildings. The proposed controller is comprised of
door view and a test room, and includes a floor plan for a single
on–off control combined with a fuzzy algorithm. On–off control
unit. Fig. 4 shows the building’s heating system and experimental
has the advantage of hardware simplicity, and fuzzy theory allows
apparatus, including an electric heater, valves, pipes, hot water
the controller to successfully adapt to the environment and
header, controller, DAQ device, and monitoring system. The system
improves overall system performance. Moreover, solar energy
components were divided into hardware, interface, and software
can be compensated for by converting the output membership
elements, as listed in Table 1. The hardware consisted of the distri-
function of the fuzzy structure.
bution system, the control system, and the heat source. The inter-
To validate the proposed controller, both simulations and
face was composed of the sensor and the DAQ device. The LabVIEW
experiments were carried out in which the proposed controller
2009 software platform was utilized to implement both the pro-
was compared to a conventional on–off controller. The simulations
posed controller and the entire operating system of the Green
allowed direct comparison between the proposed and conven-
Home Plus building.
tional designs under identical conditions, and the experiments
were conducted in actual residential buildings. The simulations
and experiments verified not only the performance of the proposed
controller but also its ability to compensate for solar irradiation.
Based on the results, the proposed method is expected to provide
better control performance than a conventional on–off controller.

2. Description of hydronic radiant floor heating systems

2.1. Construction and control types of hydronic radiant floor heating


systems

A Hydronic radiant floor heating (HRFH) system is a type of RFH


system in which hot water is used to transfer heat from a boiler to
rooms. An HRFH system consists of a heat source, a distribution
piping system, thermal space, and the control system (Fig. 1).
The heat source in HRFH systems is usually a boiler or hot water
heater powered by natural gas, oil, electricity, or solar energy, or
another comparable energy source. Heat is distributed by means
of hot water circulation through pipes that are installed in a con-
crete slab under the floor in the thermal space. Therefore, temper-
ature control in HRFH systems is carried out by controlling the
circulation or temperature of the hot water [14,15].
An HRFH system includes data flows between components
including hardware, interface, and software (Fig. 2). Provided that
sensors measure the temperature of the thermal space, the data
acquisition (DAQ) device accumulates temperature data. After that,
data are passed to the PC, which generates the control signal by
using the fuzzy algorithm. The control signal actuates valves and
pumps in the hardware components.
HRFH systems generally use one of two types of control: control
of the supply water temperature or control of the supply flow rate.
In supply water temperature control, the flow rate is constant and
the water temperature is varied; oppositely, in supply flow rate
control the water temperature is constant and the water flow rate
is varied. There are two types of supply flow rate controls: on–off Fig. 3. The Green Home Plus test house: (a) outdoor view, (b) indoor test rooms, (c)
control and variable flow rate control. In the former, a unit floor plan.
C.-S. Kang et al. / Applied Energy 155 (2015) 270–283 273

Fig. 4. Photos of heating system and experimental apparatus: (a) electric heater, (b) distribution pipes, (c) relays, (d) hot water header, (e) DAQ device, (f) monitoring system.

Table 1 3.1. Fuzzy logic


System components.

Component Item Specification Fuzzy logic is a set of mathematical principles for representing
Hardware Distribution Two-way valve HMV-415
knowledge based on degrees of membership. A fuzzy set A in a
system universe of discourse U that contains all the possible elements of
Circulation pump Flow rate: 8.3 L/min concern in each particular context or application is characterized
Control system Actuator On–off type by a membership function lA(x), which takes values in the interval
Valve controller HHC-707
[0,1]. Thus, as the value of lA(x) approaches unity, the higher the
Heat source Electric heater Capacity: 30 kW
Supply water 60 °C grade of membership function of x in A [16]. A fuzzy set A in U
temperature can be formally represented as a set of ordered pairs of a generic
Sensor Temperature sensor T-type element x and its membership value; namely,
thermocouple
Interface DAQ device DAQ board NI CDAQ A ¼ fðx; lA ðxÞÞjx 2 Ug: ð1Þ
D/A converter NI CDAQ
A/D converter NI CDAQ Classical set theory is concerned with crisp (i.e., non-fuzzy) sets;
Software Controller program LabVIEW 2009 the crisp set A of U is defined based on the function fA(x), which is
DAQ program LabVIEW 2009
called the characteristic function of A:
Operating system Windows XP

1; if x 2 A
f A ðxÞ : U ! f0; 1g; where f A ðxÞ ¼ : ð2Þ
0; if x R A
3. Design of the proposed control
By contrast, in fuzzy theory, the fuzzy set A of universe U is defined
by the function lA(x), which is called the membership function of
The design strategy for the proposed controller was divided into
set A:
two parts: the design of the mode selector and the fuzzy-based
8
controller. The mode selector chooses whether solar energy com-
< 1;
> if x is totally in A
pensation should be performed, according to the time of day. lA ðxÞ : U ! ½0; 1; where lA ðxÞ ¼ 0; if x is notin A
When the solar energy mode is enabled, the weight of the output >
:
0 < lA ðxÞ < 1; if x is partly in A
membership function is varied to deal with the effects of sunlight;
ð3Þ
otherwise, the normal mode is used, in which no solar compensa-
tion is performed. Once the mode is determined, the fuzzy-based The most commonly used fuzzy inference technique is the
controller calculates the operating duration depending on the Mamdani method. Mamdani-style fuzzy inference is performed
mode. in four steps (Fig. 5): fuzzification of the input variables (numeric
274 C.-S. Kang et al. / Applied Energy 155 (2015) 270–283

to linguistic), rule evaluation, aggregation of rule outputs, and 3.2. Design of the fuzzy-based control
defuzzification (linguistic to numeric). In the fuzzification step,
the crisp inputs are evaluated to determine their degrees of To keep the room temperature within a temperature range
belonging to relevant fuzzy sets. In the rule evaluation step, fuzzi- between set overheating and underheating thresholds, on–off
fied inputs lA(x) are applied to the antecedents of the fuzzy rules. actions should be performed before the room temperature reaches
In the aggregation step, the outputs of all rules are unified. In the either threshold. Fig. 6 shows the theoretical performance of an
defuzzification step, linguistic values are transformed to numeric anticipative control method compared with that of an on–off con-
values by means of a method such as centroid or trol method [10]. Considering the characteristics of an HRFH sys-
center-of-gravity defuzzification. tem, it is very important to decide the operation time of the

Fig. 5. Configuration of a fuzzy system, including a fuzzifier and defuzzifier.

Fig. 6. Comparison of air temperature profiles between anticipative and on–off control.

Fig. 7. Overall control logic of a RFH system including the proposed controller.
C.-S. Kang et al. / Applied Energy 155 (2015) 270–283 275

Fig. 8. Membership functions for (a) e[k], (b) G[k], (c) operating time in normal mode.

controller. Commonly, three quantities are used to decide the out-


Table 2 put response in an HRFH system: the temperature difference
Fuzzy lookup table. between a set point and the current temperature, the rate of tem-
G[k]
perature change, and the magnitude of the control input.
Unfortunately, in the building environment studied herein, the
NNB NB NM NS ZO PS PM PB PPB
magnitude of the control input was fixed during the unit time
e[k] L Bl Bl Bl Bl Bl Bl Bl Bl Sl because the on–off control was embedded and the hot water tem-
M Bl Bl Bl Bl Bl Bm Sm Bs Ss
perature was constant. Thus, the operating time of the controller
H ZO ZO ZO ZO ZO ZO ZO ZO ZO
was considered instead of the magnitude of the control input. In
Bl: 100 min, Sl: 90 min, Bm: 70 min, Sm: 60 min, Bs: 40 min, Ss: 30 min, ZO: 0. other words, a longer operating time is necessary if the difference
between the current room temperature and the overheating
threshold is large; oppositely, when the rate of room temperature
change is high, the room temperature is likely to reach the heating
limit quickly, and thus it is necessary to turn off the boiler earlier,
before the room temperature reaches the overheating threshold (or
to turn it on sooner, for the underheating case).
In the proposed scheme, the overall system comprises the mode
selector, fuzzy-based controller, the temperature sensor, and heat-
ing systems including pumps, pipes, a boiler, and so on (Fig. 7). The
mode selector has two modes: normal mode and solar energy com-
pensation mode. Because solar illumination strongly impacts the
temperature of interior room air during a certain time of day,
approximately from 9 a.m. to 12 noon, the proposed controller
was designed to include a mode in which control was adapted to
compensate for temperature variations due to sunlight. In the pro-
posed controller, the on–off signals are determined by using the
fuzzy-based controller, which takes two inputs and generates
one output. One of the two inputs is the difference between the
set point and the current room air temperature, and the other is
the rate of temperature change. The output is the operation dura-
Fig. 9. Membership functions for operating time in solar energy compensation tion, which represents the decision of whether to turn the boiler on
mode. or off. Because the room temperature was measured at one-minute
276 C.-S. Kang et al. / Applied Energy 155 (2015) 270–283

Fig. 10. Schematic diagram of heat flow affecting room air temperature [13].

and ZO approximately zero; these abbreviations are combined to


represent different cases; for example, PB stands for big positive,
NS for small negative, and so on. The fuzzy set Cl is shown in Fig. 8c.
To properly adjust the boiler operating time, a set of rules like
those shown in Table 2 may be employed. In Table 2, B and S
respectively represent big and small, and l, m, s and ZO respectively
represent long, medium, short, and no operation. Thus, Bl stands
for big-long, Ss for small-short, and so on. The general fuzzy logic
system with a singleton fuzzifier, a product inference engine, and
a center average defuzzifier can be designed as [18]:

P27
l¼1 ðy lA ðe½kÞlB ðG½kÞÞ
l l l
f ðxÞ ¼ P27 l ; ð5Þ
l¼1 lA ðe½kÞlB ðG½kÞ
l

where f(x) represents the duration of operation. The fuzzy sets Al


and Bl are shown in Fig. 8a and b, respectively, and y
l is the center
Fig. 11. Test model for radiant heat analysis. of the corresponding fuzzy sets, shown in Fig. 8c.

intervals in the test facility, the on–off operating state is deter-


3.3. Solar energy compensation mode
mined every minute in the proposed controller.
In the proposed controller, the operating time is determined
In well-insulated buildings, large variations in room tempera-
based on the difference between the set point and the current
ture can occur because of changes in internal loads such as resident
room air temperature, e[k], and the rate of temperature change,
occupancy, appliances, light bulbs, and solar gain. Among these
G[k]. More specifically, the operating time is determined by a set
factors, solar gain highly affects room temperature. This fact has
of fuzzy rules in the following form [17].
been verified by previous experiments including the use of on–
off controllers. Especially, from 9 a.m. to 12 noon, solar energy is
If e½k is Al and G½k is Bl ; then the operating time is C l ð4Þ
the variable that affects the control performance most strongly.
Thus, when needed, the algorithm presented herein compensates
Here, Al, Bl, and Cl are fuzzy sets in the corresponding supporting
for the effect of direct sunlight when determining the system oper-
sets. Note that the fuzzy rules in Eq. (4) were found using heuristic
ation. Unlike in normal operation, to consider the degree of solar
measures. The membership function for these fuzzy sets for e[k] is
gain, the output membership function for the operating time must
shown in Fig 8a; in this figure, L stands for low, M for middle, and
be transformed by adding a weight factor w. This membership
H for high. The fuzzy set Bl is characterized by the membership
function is shown in Fig. 9.
functions shown in Fig. 8b, where B stands for big, M medium, S
The fuzzy-based controller represented by Eq. (6) operates from
small, N negative, NN very negative, P positive, PP very positive,
9 a.m. to 12 noon, and the fuzzy-based controller represented by
Eq. (5) operates otherwise.
Table 3
Material properties of a floor and ceiling. P27
l¼1 ðyC lA ðe½kÞlB ðG½kÞÞ
 l l l

Material Thickness (m) Specific heat (kJ/kg K) Density (kg/m3)


f ðxÞ ¼ P27 l ð6Þ
l¼1 lA ðe½kÞlB ðG½kÞ
l

Cement mortar 0.04 0.89 2000


Foamed concrete 0.05 0.84 350 Here, f⁄(x) represents the operating time. The fuzzy sets Al and Bl
Insulation 0.025 1.42 28 are shown in Fig. 8a and b, respectively, and ylC is the center of the
Concrete 0.21 0.92 2200
corresponding fuzzy sets, shown in Fig. 9.
C.-S. Kang et al. / Applied Energy 155 (2015) 270–283 277

8
4. Development of thermal analysis program considering solar >
> q1 ¼ 80½W=m2   A1
>
>
gain >
>
>
> q2 ¼ 2:41A1 ðT 1  T a Þ1:31 =D0:08
e
>
>
>
>
A thermal analysis program that considers solar gain was devel- >
>
> q3 ¼ 1:87A6 ðT a  T 6 Þ1:32 =H0:05
3
>
>
oped to verify the validity of the proposed controller. The thermal >
> q ¼ 1:87A8 ðT a  T 8 Þ1:32 =H0:05
>
>
analysis program was based on a program proposed by Ahn and > 4
>
8
>
Song [13] as a simulation tool to test controllers for HRFH systems.
>
>
> q5 ¼ 1:87A7 ðT a  T 7 Þ1:32 =H0:05
7
>
>
>
>
This program is very useful to verify various controllers’ perfor-
< q6
> ¼ 1:87A5 ðT a  T 5 Þ1:32 =H0:05
5
mance under consistent conditions of outdoor temperature, solar ð7Þ
> q7 ¼ U 6 A6 ðT 6  T out Þ
irradiation, building materials, etc.; however, it does not include >
>
>
> q8 ¼ U 8 A8 ðT 8  T out Þ
the solar energy model that is the main issue under consideration >
>
>
>
in this paper. This section presents the modifications that were >
> q9 ¼ U 7 A7 ðT 7  T out Þ
>
>
made to this thermal analysis program to consider solar gain. >
>
>
> q10 ¼ U 10 A5 ðT 5  T out Þ
>
>
>
> q ¼ 1:87A ðT  T Þ1:32 =H0:25
4.1. Thermal analysis program >
>
>
> 11 2 2 a 2
>
>
>
> q ¼ 1:87A ðT  T Þ 1:32
=H 0:05
>
> 12 3 a 3
Fig. 10 illustrates three kinds of heat flow within a room. The >
:
room has six walls: one floor, one ceiling, three internal walls, q13 ¼ 1:87A4 ðT a  T 4 Þ1:32 =H0:05
and one external wall including one window and one door Here, qi denotes the heat transfer rate of node i, Ti is the tempera-
(Fig. 11). Among them, the floor and the ceiling are multi-layer ture of node i, Ta is the room air temperature, De is the equivalent
walls whose material properties are listed in Table 3. It was diameter of the floor, and H is the wall height, respectively. For cal-
assumed that heat was supplied by circulation of hot water in culating radiant heat transfer rates, enclosure analysis methods
the pipes embedded in the floor as well as by solar irradiation. were used as follows [20,21].
Heat transfer comprises conduction heat exchange between the
X
n
surface of a wall and its interior, convective heat transfer between qrk ¼ Ak  ek  r  T 4k  Aj  ej  r  T 4j  Gjk
the surface of a wall and indoor air, and radiant heat exchange j¼1
between the surface of a wall and the surfaces of other walls
Gjk ¼ F jk  ek þ F jk  q1  G1k þ . . . þ F jn  qn  Gnk ; j ¼ 1; 2; . . . ; n
[19]. The heat transfer rates by convection and conduction were
calculated as follows. ð8Þ

Table 4
Parameters used in the simulation study.

Parameter Description Value Parameter Description Value


2
Ca (kJ/K) Indoor air heat capacity 47.6468 A1 (m ) Floor dimension 14.8
C1 (kJ/K) Floor heat capacity 7576.6 A2 (m2) Ceiling dimension 14.8
C2 (kJ/K) Ceiling heat capacity 7576.6 A3 (m2) Wall 1 dimension 10.4
C3 (kJ/K) Wall 1 heat capacity 639.2 A4 (m2) Wall 2 dimension 10.4
C4 (kJ/K) Wall 2 heat capacity 478.4 A5 (m2) Wall 3 dimension 4.7
C5 (kJ/K) Wall 3 heat capacity 681.15 A6 (m2) Wall 4 dimension 9.1
C6 (kJ/K) Wall 4 heat capacity 411.52 A7 (m2) Window dimension 4.4
C7 (kJ/K) Window heat capacity 120.67 A8 (m2) Door dimension 2.4
C8 (kJ/K) Door heat capacity 411.52 H (m) Wall height 3.5986
De (m) Equivalent diameter 3.8243 H2 (m) Wall height 2.5288
r (kJ/K) Boltzmanm constant 1.3805  1020 H3 (m) Wall height 4.1127
H5 (m) Wall height 4.1127 H7 (m) Wall height 4.1127
H8 (m) Wall height 4.1127

Table 5
Shape factors of surfaces for radiant heat transfer analysis.

Floor Ceiling Wall 1 Wall 2 Wall 3 Wall 4 Window Door


Floor 0.0000 0.5622 0.1328 0.1328 0.0480 0.1011 0.0676 0.0154
Ceiling 0.5622 0.0000 0.1328 0.1328 0.0480 0.1045 0.0676 0.0120
Wall 1 0.4138 0.4138 0.0000 0.0584 0.0245 0.1035 0.0378 0.0082
Wall 2 0.4138 0.4138 0.0584 0.0000 0.0245 0.1036 0.0378 0.0082
Wall 3 0.3496 0.3496 0.1781 0.1781 0.0000 0.0040 0.0000 0.0008
Wall 4 0.3923 0.4055 0.1288 0.1288 0.0021 0.0000 0.0025 0.0000
Window 0.4469 0.4469 0.0803 0.0803 0.0000 0.0050 0.0000 0.0006
Door 0.4967 0.3881 0.0851 0.0851 0.0017 0.0000 0.0033 0.0000

Table 6
Characteristic values of surfaces for radiant heat transfer analysis.

Floor Ceiling Wall 1 Wall 2 Wall 3 Wall 4 Window Door


Emissivity 0.70 0.72 0.72 0.72 0.72 0.72 0.74 0.75
Reflectivity 0.25 0.27 0.27 0.27 0.27 0.27 0.05 0.23
Dimension 14.8 14.8 10.4 10.4 4.7 6.7 4.4 2.4
278 C.-S. Kang et al. / Applied Energy 155 (2015) 270–283

Here, the subscripts j and k denote each surface, Ak denotes the sur- program was useful for developing a controller that considered
face dimension, Tk the surface temperature, r the Boltzmanm con- solar energy and HRFH systems.
stant, ek the emissivity, qk the reflectivity, and Fjk the shape factor.
Heat transfer characteristics were analyzed by using the follow- 4.2.1. Modeling of solar irradiation
ing equivalent resistance–capacitance circuit analysis method For modeling of solar irradiation, the exact position of the sun in
[22,23]. relation to the building is very important. There is a well-known
8 dT a procedure to obtain the exact location, consisting of four steps:
>
> C a dt ¼ q2  q12  q3  q4  q5  q6  q13  q14
>
> obtaining the apparent solar time (AST) and hour angle (H), calculat-
>
> dT 1
> C 1 dt ¼ q1  q2  qr1
>
>
ing the solar declination (d), and deriving the altitude and azimuth
>
> dT angle of the sun [24].
>
> C 2 dt2 ¼ q12  qr2
>
> The first step is to obtain the AST, which is based on the appar-
>
>
> C 3 dTdt3 ¼ q13  qr3
>
< ent motion of the actual sun. The AST is derived as follows:
C 4 dTdt4 ¼ q14  qr4 ð9Þ
>
> AST ¼ LST þ ET=60 þ ðLON  LSMÞ=15 ð10Þ
>
>
>
> C 5 dTdt5 ¼ q6  q10  qr5
>
> In this equation, the terms LST, ET, LON, and LSM respectively
> C dT 6 ¼ q  q  q
>
>
> 6 dt 3 7 r6
indicate local standard time, equation of time, longitude of site, and
>
> dT 7
>
> C ¼ q  q  q
>
> 7 dt 5 9 r7 longitude of local standard time. ET describes the discrepancy
>
:
C 8 dTdt8 ¼ q4  q8  qr8 between two kinds of solar time: AST and mean solar time (MST).
The MST is a time track that refers to a fictitious ‘mean sun’ with
The parameters in Eqs. (7)–(9) are given in Table 4 [24]. Note noons 24 h apart. Based on these two kinds of solar time, ET is
that walls 1, 2, and 4 were internal walls and that wall 3 was an approximated as follows [25]:
outside wall. The shape factors between each wall that were used
in Eq. (8) are listed in Table 5. The characteristic values of emissiv- ET ¼ 2:2918½0:0075 þ 0:1868 cosðCÞ  3:2077 sinðCÞ
ity, reflectivity, and dimension for each surface are listed in Table 6.  1:4615 cosð2CÞ  4:089 sinð2CÞ; ð11Þ

4.2. Modified thermal analysis program where

n1
Variations of solar radiation according to the time of year and C ¼ 360 ð12Þ
time of day are one of the most important factors determining 365
the indoor air temperature of an HRFH system, because the build- and n indicates the day of the year.
ing envelope receives solar radiation differently depending on the As the second step, the obtained AST is converted into an hour
sun’s position. However, the conventional thermal analysis pro- angle, defined as the angular displacement of the sun east or west
gram proposed by Ahn and Song [13] does not include a solar of the local meridian due to Earth’s rotation, and expressed in
model. In this section, a modified thermal analysis program is pro- degrees as
posed that was capable of verifying the validity of a controller that
included a solar energy compensation algorithm. The modified H ¼ 15ðAST  12Þ: ð13Þ

Fig. 12. Solar angles for vertical and horizontal surfaces.


C.-S. Kang et al. / Applied Energy 155 (2015) 270–283 279

28
On-Off Control
Proposed Control
27

26

25

Temperature (oC) 24

23

22

21

20

19

18
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
Time (min) x 104

Fig. 13. Simulation result: comparison between the proposed control and on–off control.

(a) RMSE (b) Heating time


Fig. 14. Simulation results with respect to error and heating time.

Table 7 1
b ¼ sin ðcos L cos d cos H þ sin L sin dÞ ð15Þ
Experimental conditions.
The solar azimuth angle / is defined as the angular displacement
Control Set point and Supply water Period
method differential temperature
between south and the projection of the Earth–Sun line on the hor-
izontal plane; it is determined as follows.
On–off control 20 ± 0.5 °C 60 °C 24 h
Proposed 20 °C 60 °C 24 h
/ ¼ cos1 fðcos H cos d sin L  sin d cos LÞ= cos bg ð16Þ
control
In addition to information on the sun’s direction relative to the
building, knowledge of the receiving surface is necessary to calcu-
Then, as the next step, the solar declination should be consid- late the amount of solar radiation that will affect the building’s
ered; this is the angle between the Earth–Sun line and the equato- indoor room temperature. The orientation of a receiving surface
rial plane. Solar declination varies throughout the year because is usually characterized by its tilt angle R and surface azimuth w:
Earth’s equatorial plane is tilted at an angle 23.45° to the orbital the tilt angle is the angle between the surface and the horizontal
plane. This variation not only causes the changing seasons, with plane; and the surface azimuth is defined as the displacement from
their unequal periods of daylight and darkness, but also affects south of the projection, on the horizontal plane, of the normal to
the solar altitude. The solar declination d is given as follows. the surface.
  So far, two kinds of information have been derived: solar alti-
n þ 284 tude and azimuth related to the direction of the sun relative to
d ¼ 23:45 sin 360 ð14Þ
365 the building, and information on the receiving surface illuminated
Finally, the solar altitude and azimuth can be calculated by by the sun. By using this information, the angle of incidence h,
using three terms: the hour angle H, the solar declination d, and meaning the angle between the line normal to the irradiated sur-
the local latitude L. Fig. 12 illustrates the use of the solar altitude face and the Earth–Sun line, can be calculated as follows.
above the horizontal and the solar azimuth measured from the h ¼ cos1 ðcos b cos c sin R þ sin b cos RÞ ð17Þ
south as a means of expressing the sun’s position. The solar altitude
angle b is defined as the angle between the horizontal plane and Here, the surface–solar azimuth angle c is defined as the angular dif-
the Earth–Sun line; it is determined as follows. ference between the solar azimuth / and the surface azimuth w:
280 C.-S. Kang et al. / Applied Energy 155 (2015) 270–283

Fig. 15. Experimental condition: outdoor temperature and solar irradiation.

(a) RMSE (b) Heating time

(c) Positive error max (d) Negative error max


Fig. 16. Error and heating time during the experiments.

8 dT
c¼/w ð18Þ >
> C a dta ¼ q2  q12  q3  q4  q5  q6  q13  q14
>
>
>
> dT 1
Because the angle of incidence affects the intensity of the direct >
> C 1 dt ¼ q1  q2  qr1 þ I
>
>
component of solar radiation striking the surface, calculating the >
> dT 2
>
> C 2 dt ¼ q12  qr2
angle of incidence is a very important procedure to exactly deter- >
>
>
>
mine the solar irradiation received by the building. >
> dT 3
> C 3 dt ¼ q13  qr3
>
<
C 4 dTdt4 ¼ q14  qr4 ð19Þ
4.2.2. Modeling of an HRFH system including solar energy >
>
>
>
In this subsection, we propose a novel HRFH system model that > C 5 dTdt5 ¼ q6  q10  qr5
>
>
>
considers solar energy; this model was used to assist the develop- >
> dT 6
>
> C 6 dt ¼ q3  q7  qr6
ment of a controller that considers solar gain. In contrast to the >
>
>
>
conventional HRFH system represented by Eq. (9), the proposed >
> dT 7
> C 7 dt ¼ q5  q9  qr7
>
>
HRFH system given in Eq. (19) involves one solar energy term, >
: dT 8
under the assumption that solar energy affects only the floor.
C 8 dt ¼ q4  q8  qr8
C.-S. Kang et al. / Applied Energy 155 (2015) 270–283 281

4.3. Simulation results 5. Experiment: application to the Green Home Plus building

In our simulations, we assumed that the building surface was 5.1. Experimental conditions
perpendicular to the ground, i.e., that the tilt angle R was 90° and
that it faced south, i.e., that the surface azimuth w was 0°. Typical Table 7 lists the conditions during the in-building experiments
weather data for winter in the Republic of Korea was used for using the two different controls. The supply water temperature
the simulations. The duration of 30 days was simulated to include was set to 60 °C for both controls. The experiment was conducted
various weather situations, the desired room temperature was set for 24 consecutive hours; however, the first three hours of both
to 20 °C, and both a conventional on–off controller and the pro- periods were used to allow the experimental space to reach similar
posed controller were simulated. The results clearly showed that initial temperatures near 20 °C. The operating band of the on–off
the conventional on–off controller varied from the desired temper- control was ±0.5 °C; in other words, when the measured tempera-
ature more than the proposed controller, meaning that the pro- ture fell to 19.5 °C, the controller directed the boiler system to sup-
posed controller effectively anticipated the on–off timing of the ply hot water, and when the measured temperature reached
boiler by means of fuzzy logic (Fig. 13). In the figure, peaks during 20.5 °C, the controller directed the boiler system stop supplying
the daytime operation periods indicate the control response. Some hot water. The fuzzy-based control did not need operating band,
temperature variations were inevitable because of the high solar but only a set point temperature; under this control scheme, the
gain; however, the variations were greatly reduced when the solar supply of hot water was started or stopped by considering the ten-
energy compensation method was used. dency of temperature changes in addition to the current
To emphasize the advantages of the proposed controller, the temperature.
root mean square error (RMSE) and heating times of the two con- The on–off control and the proposed fuzzy-based control were
trollers were analyzed. The RMSE of the proposed controller is each tested on winter days in Korea. Due to the project schedule,
about 0.4 °C smaller than that of the conventional on–off controller the on–off control experiment was conducted in January whereas
(Fig. 14a). Under operation by the proposed controller, the boiler the fuzzy-based control experiment was conducted in March.
operated for less time than when the conventional on–off con- Hence, there were differences in outdoor temperature and solar
troller was used. In summary, the proposed controller was both irradiation between the two days of the experiments (Fig. 15).
more accurate and more energy-efficient. However, the validity of the proposed controller had been verified

Fig. 17. Experimental temperature variations and on–off timing during the use of the on–off controller, in (a) room 1, (b) room 2, (c) room 3, and (d) room 4.
282 C.-S. Kang et al. / Applied Energy 155 (2015) 270–283

by means of the simulations conducted under the same conditions, break down, a minimum operation time was applied to prevent
as described in the previous section. In this section, we discuss the malfunctions. Overall, the proposed method controlled the room
experimental results, which confirmed the applicability of the pro- temperature well during the experiment.
posed controller in the actual environment of the Green Home Plus By contrast, the performance of the on–off control was much
building. worse, including large overshoots that resulted in energy losses
(Fig. 17). There were larger overshoots during certain time inter-
5.2. Analysis of experimental results vals; specifically, overshoot from 9:00 am to 12:00 noon was the
most severe. This tendency was observed in each of the rooms.
Experimental room temperature variations in four rooms dur- As previously discussed, the proposed control was designed to
ing the use of the on–off control and the proposed control are compensate for this phenomenon by the inclusion of a weighting
shown in Figs. 17 and 18, respectively. Note that the room temper- factor. Consequently, when the proposed solar energy compensa-
ature during the use of the on–off control initially start above tion method was used, the largest overshoots were significantly
20 °C, whereas it was below 20 °C for the proposed control. This reduced (Fig. 18). Moreover, there was no undershooting except
discrepancy occurred because the experiments were conducted once in room 4. This result indicates that the fuzzy-based control
on different days. In the on–off control case, the boiler was turned directed the system to turn on at the proper times.
on the day before the experiment, whereas it was turned off prior Fig. 16 presents the experimental results of error and heating
to the experiment in the fuzzy-based control case. time for each room. For instance, when the on–off control was
Further, recall that the on–off control turns the boiler on or off used in room 1, the RMSE was 1.047, and the error ranged from
based only on the current temperature—on at 19.5 °C and off at 0.75 to 3.49, where as in the case of the fuzzy-based control
20.5 °C—whereas the proposed control determines the on–off tim- the RMSE was 0.489 and the error ranged from 0.64 to 1.16.
ing by considering factors such as the current room temperature, That is to say, less underheating and overheating occurred when
previous room temperatures, and solar irradiation. As shown in the proposed method was used. Moreover, the heating time under
Fig. 18, the on and off times occurred before the room temperature the proposed controller was less than that under the on–off con-
reached 19.5 or 20.5 °C, respectively. Thus, the amounts of over- troller. Similar results were observed in the other rooms. In con-
shoot and undershoot were reduced by employing the proposed clusion, the proposed method had better control performance and
control. Because frequent on–off signals may cause equipment to saved energy.

Fig. 18. Experimental temperature variations and on–off timing during the use of the proposed controller, in (a) room 1, (b) room 2, (c) room 3, and (d) room 4.
C.-S. Kang et al. / Applied Energy 155 (2015) 270–283 283

6. Conclusions [3] Bi Q, Cai W-J, Wang Q-G, Hang C-C, Lee E-L, Sun Y, et al. Advanced controller
auto-tuning and its application in HVAC systems. Control Eng Practice
2000;8:633–44.
In this paper, a fuzzy logic-based advanced on–off control [4] Borresen B, Grindal A. Controllability—back to basics. ASHRAE Trans Res
design for HRFH systems was proposed to improve control perfor- 1990:817–9.
[5] Geng G, Geary G. On performance and tuning of PID controllers in HVAC
mance and reduce energy loss compared to a conventional on–off
systems. Control Applications, 1993, Second IEEE Conference on: IEEE; 1993. p.
control method. The proposed controller was designed to take into 819–24.
consideration the current temperature state and the trend of tem- [6] Kawashima M, Dorgan CE, Mitchell JW. 3849 Hourly thermal load prediction
for the next 24 Hours by ARIMA, EWMA, LR, and an Artificial Neural Network.
perature change. Additionally, because solar gain during the late
ASHRAE 1995;101:186–200.
morning greatly affects the system performance, a solar energy [7] Kreider J, Claridge D, Curtiss P, Dodier R, Haberl J, Krarti M. Building energy use
compensation method was included in the proposed controller. prediction and system identification using recurrent neural networks. J
The proposed fuzzy-based control combines the hardware simplic- SolEnergy Eng 1995;117:161–6.
[8] Sierra E, Hossian A, Britos P, Rodriguez D, Garcia-Martinez R. Fuzzy Control for
ity of an on–off control system and the improved anticipatory con- improving energy management within indoor building environments,
trol performance of fuzzy theory. Electronics, Robotics and Automotive Mechanics Conference, 20 07 C ER M A
A novel program for thermal analysis of HRFH systems was 20 07, Springer, 2007. p. 412–6.
[9] Lee J-Y, Yang I-H, Song S-Y, Kim H-S, Kim K-W. A study of the predictive control
developed to include consideration of solar energy; this program of the ondol system in apartments. Proc Build Simul 1999:1.
was used to verify the validity of the proposed controller by com- [10] Lee J-Y, Yeo M-S, Kim K-W. Predictive control of the radiant floor heating
paring it to the conventional on–off control under identical condi- system in apartment buildings. J Asian Archit Build Eng 2002;1:105–12.
[11] Argiriou AA, Bellas-Velidis I, Balaras CA. Development of a neural network
tions of outdoor temperature, solar irradiation, and building heating controller for solar buildings. Neural Networks 2000;13:811–20.
materials. The simulation results showed that the proposed con- [12] Argiriou AA, Bellas-Velidis I, Kummert M, André P. A neural network controller
troller was better than a conventional on–off controller in terms for hydronic heating systems of solar buildings. Neural Networks
2004;17:427–40.
of reducing overheating and underheating.
[13] Ahn B-C, Song J-Y. Control characteristics and heating performance analysis of
As validated by our experimental results, the proposed con- automatic thermostatic valves for radiant slab heating system in residential
troller more effectively determined the proper on–off timing than apartments. Energy 2010;35:1615–24.
[14] Baek J, Choi J-y, Jang S-Y, Park M. Simulational analysis of adaptive outdoor
did the system using only the on–off controller. Proper on–off tim-
reset control based on a fuzzy target temperature gap for a hydronic radiant
ing reduced overheating and underheating, thereby diminishing floor heating system. J Asian Archit Build Eng 2010;9:251–7.
energy consumption. Furthermore, the overshoot resulting from [15] Dodge RW. Radiant floor heating. New York: McGraw-Hill Book Company;
midday solar gain was considerably reduced. It is expected that 1999.
[16] Zadeh LA. Fuzzy sets. Inf Control 1965;8:338–53.
the results of this study will contribute to reducing energy waste [17] Zhao Z-Y, Tomizuka M, Isaka S. Fuzzy gain scheduling of PID controllers,
and improving thermal comfort in building applications. systems, man and cybernetics. IEEE Trans 1993;23:1392–8.
[18] Wang L-X. A course in fuzzy systems. USA: Prentice-Hall press; 1999.
[19] Ma P, Wang L-S, Guo N. Modeling of hydronic radiant cooling of a thermally
Acknowledgements homeostatic building using a parametric cooling tower. Appl Energy
2014;127:172–81.
This research was supported by a grant (11 High-tech Urban [20] Siegel R, Howell JR. Thermal radiation heat transfer. National Aeronautics and
Space Administration, Cleveland, OH (United States). Lewis Research Center;
G03) from High-tech Urban Development Program funded by 1992.
Ministry of Land, Infrastructure and Transport of Korean govern- [21] Ahn B. Optimal operation methods for hot water radiant panel heating system
ment. The authors would like to thank Myoung Souk Yeo and in residential buildings. In: Proceedings of the SAREK 2004 summer annual
conference 2004. p. 800–6.
Hee Gang Kim for their advice regarding the development of the
[22] Lü X, Lu T, Kibert CJ, Viljanen M. A novel dynamic modeling approach for
thermal analysis program to consider solar gain, and gratefully predicting building energy performance. Appl Energy 2014;114:91–103.
acknowledge the Editor, Associate Editor, and anonymous review- [23] Dovrtel K, Medved S. Weather-predicted control of building free cooling
system. Appl Energy 2011;88:3088–96.
ers for their invaluable comments and suggestions.
[24] Committee AS. ASHRAE HANDBOOK: Fundamentals 2013. ASHRAE; 2013.
[25] Iqbal M. An introduction to solar radiation. Elsevier; 2012.
References

[1] Olesen BW. Radiant floor heating in theory and practice. ASHRAE J
2002;44:19–26.
[2] Zhai X, Yang J, Wang R. Design and performance of the solar-powered floor
heating system in a green building. Renewable Energy 2009;34:1700–8.

You might also like