Supervisor of Digital PI-like Fuzzy
Logic Controllers for Indoor
Lighting Control in Buildings
K. Alexandridis and A. I. Dounis
Department of Automation, Technological Educational Institute of Piraeus,
Piraeus, P. Ralli & Thivon 250, Greece, Tel: 2105381338, email:
aidounis@otenet.gr.
Abstract
In this paper, we develop a supervisor of digital PI-like fuzzy logic controllers
(FLC) for indoor lighting conditions control in buildings. The proposed fuzzy
control system has hierarchical structure. This structure consists from one
supervisor and two fuzzy logic controllers. The supervisor evaluates the
daylight and artificial lighting and decides by logic-based switching for the
fuzzy controllers’ operation. The structure of a PI-like fuzzy logic controller is
presented.
The control system is implemented in a simulation environment including
reference models for the building. The environment combines TRNSYS
(Transient System Simulation Program) and MATLAB software’s. Τhe role of
the real system is played by a model implemented in TRNSYS. The control
system is implemented in MATLAB. The communication between TRNSYS
and MATLAB is realized by a TRNSYS TYPE that calling the MATLAB Engine
Library.
The simulation results show that the proposed fuzzy control system
successfully manages the illuminance comfort and the energy conservation.
Keywords: Digital Fuzzy Logic Controller, Energy Saving, Lighting comfort,
Building, Supervisor.
1. Introduction
The problem of energy saving and the achievement of visual comfort
conditions in the interior environment of a building is multidimensional.
Scientists from a variety of fields have been working on it for quite a few
decades, but it still remains an open problem. People spend about 80% of
their lives inside buildings. So, achieving lighting comfort conditions in a
building is very important and has direct implication to the productivity of the
occupants and indirect implication to the energy efficiency of the building.
Indoor lighting in buildings is a topic of a major importance for researchers.
Dounis [4] proposed a fuzzy control scheme for visual comfort in a building
zone. The indoor illuminance levels together with the Daylight Glare Index are
taken into account by the fuzzy control scheme to regulate the shading and
electric lighting [7].
User behaviours concerning the blind position are often very complex and
hardly predictable. Guillemin and Moltemi [6] used Genetic Algorithms in a
shading-device controller with goal to learn the user preferences. Guillemin
and Morel [5] presented a self-adaptive multi-controller system. In this system
every controller works in order to help the others. The overall optimization of
the system realised through the use of GA.
In the Lah’s paper [9,10] proposed a modern approach to control the inside
illuminance with fully automated fuzzy system for adjusting shades, which
responds constantly to the changes in the available solar radiation, which
makes decisions as it follows the human thinking process. The main fuzzy
logic controller is linked with an auxiliary conventional PID controller. The goal
of this lower level controller is the control the roll position.
Hybrid systems like ANFIS (Adaptive Neurο-Fuzzy Inference System) have
been used for prediction and control of the artificial lighting in buildings,
following the variations of the natural lighting [8].
The present paper presented a method supervision control that uses digital
PI-like FLC to improve both lighting level and energy efficiency at the same
time. The main goal of the proposed supervisory control system is to take full
advantage of daylight for inside lighting.
2. Considered System
2.1 Simulation environment (MATLAB-TRNSYS)
This environment combines TRNSYS 16 [16] and MATLAB software (Fig. 1).
The building model is implemented in TRNSYS and the control system is
implemented in MATLAB. Simulation time step is 6 min. Controllers outputs
belong to interval [0,1] and Φi is the maximum power of each actuator. The
simulation environment, shown in Figure 1, includes the following
components:
1) TRNSYS TYPE 56 module: Multi-zone Building modeling.
2) TRNSYS TYPE 155: The interface between TRNSYS and MATLAB. The
controllers are implemented in MATLAB. For the controllers for which an
executable program is available, the file data transfer and the call to
executable routine are also implemented in MATLAB. The TYPE 155 is a
standard TRNSYS routine.
3) TYPE 9 module: This component is used to read the weather data files
(TRY is generated by meteorological data from Athens, Greece [1].
4) TYPE 16 module: This component is a radiation processor with smoothing.
5) The calculations relevant to the natural and artificial illumination, the
development of the fuzzy controllers and supervisor are implemented in the
MATLAB.
All simulations concerned a passive solar building characterized by an
important south-facing window glazed area (3m2), area 45 m2, volume 135 m3
and by a high thermal inertia, light transmittance of the window glazing mean
(τ=0.817), reflectance of all indoor surfaces (ρ=0.4). In the TRNSYS there
exist an electric lighting (10 lamps, 0-1000 lux, 800 W total), and a shading
device (curtain). The controller’s initial set point is: indoor Illuminance= {800600-500-800}lux. i is the maximum power of each actuator.
TRNSYS
TYPE 9
MATLAB
Illuminance
and CO2
Calculation
TYPE
155
TYPE 16
Φi
Controllers Supervisor
*
i
u

TYPE 56
(Actuators)
Figure 1: Simulation block diagram.
2.2 Lighting
Indoor Natural Lighting
The average indoor illuminance Εin (lx) [11] is calculated using the equation
AwEv
Ein 
(1)
Ain (1   )
where
Aw (m2) the window surface
τ (-) the light transmittance of the window glazing
Εv (lx) the vertical illuminance on the window
Ain (m2) the total area of all indoor surfaces
ρ (-) the area weighted mean reflectance of all indoor surfaces.
The vertical illuminance on the window Εv (lx) is given by the following
equation
(2)
Ev  k G Gv
with
kG (lm.W -1) the luminous efficacy of global solar radiation
Gv (W.m-2) the global solar radiation on the window surface
The luminous efficacy of global solar radiation [13] can be calculated by the
following relation
 D 
D
k G  h k D  1  h k s
(3)
Gh
 Gh 
with
Dh (W.m-2) the diffuse horizontal solar radiation
Gh (W.m-2) the global horizontal solar radiation
kD (lm.W -1) the luminous efficacy of diffuse solar radiation
kS (lm.W -1) the luminous efficacy of beam solar radiation.
The luminous efficacy of diffuse solar radiation [12] is calculated using the
equation
(4)
k D  144  29C
1  C  0.55 NI  1.22 NI 2  1.68 NI 3
Dh
Gh
NI 
1  0.12037 sin 0.82 ( z )
with θz (deg) the solar zenith angle.
1
(5)
(6)
Finally, the luminous efficacy of the beam solar radiation [2] can be calculated
using the relation
kS = 17.72 + 4.4585 θz – 8.7563 10-2 (θz)2 + 7.3948  10-4 (θz)3 – 2.167 (7)
 10-6 (θz)4 – 8.4132  10-10 (θz)5
A qualitative criterion for the control performance is the value of Illiminance
Discomfort Index (IDI) (Equation 8). K is the index of the sample, Ti the
sampling time and ei the sample error.
k
 ( ei Ti )
IDI  i1
k
 Ti
i1
(8)
Artificial Lighting
The Equation below is used to calculate the average artificial light intensity
inside the buildings:

AL

u*AL  N  ( P V  n)
2  ( H  h)2
(9)
where
u *AL : The actuating signal of the artificial light controller, ranging from 0-1. This
signal is driven by the artificial lighting fuzzy controller. The same signal is
also fed into the building model (Archimed.bui) to drive the actuator for the
artificial lights. If u*AL  0 means that all lights are off. If u *AL  1 means that all
lights are on at full power. In the latter case, the equivalent intensity is
approximately EAL=1000 lux. N: Number of light lamps (N=10), P: The power
per lamb (P=60W), V: The luminous efficacy/efficiency of each lamp (V=60
lumen/W), n: The power efficiency of each lamp (n=0.7), H: The height lamp
from floor (H=3m), h: The height reference working level, measured from the
floor (h=1m).
3. Digital PI-like FLC
The proposed PI-like FLC is useful because in the building control systems
there are actuators with continuous output such as variable speed fans, hot
water heating systems, electrical heaters, air inlets. All the membership
functions of the PI-like FLC inputs/outputs are shown in Figures 3 and 4. The
input/output normalization maps the state variables on the interval [-1,+1]. The
scaling factors are chosen to be Ge =1/1200, Ge =1/120000 and Gu =1. These
scaling factors have been found via simulations (trial and error). The output of
each controller is u(k )  {u AL , uSH } The fuzzy control rules are presented in
Table 1 and 2.
e(k ) Ge
Fuzzy
PI
e(k )
r (k )
+ Σ-
+
Σ
+ -
u (k  1)
Ge
z 1
e( k  1)
Gu u ( k )
u (k )
TRNSYS
y (k )
z 1
-
Σ
Figure 2: Digital implementation of a PI-like FLC
Membership function
1
NB
NS
NM
ZE
PS
PM
PB
0,8
NB
NS
NM
ZE
PM
PS
PB
0,6
0,4
-1
0,2
-0.5
0
+1
0.5
e, Δe
0
-1
-0,5
-0,25
0
0,25
ΔuSH, ΔuAL
0,5
1
Figure 3: Membership functions of the
FLCs output variables.
Δe
e
NB
NM
NS
ZE
PS
PM
PB
Figure 4: Membership functions of the
FLCs input variables.
NB
NM
NS
ZE
PS
PM
PB
PB
PB
PB
PB
PM
PS
ZE
PB
PB
PB
PM
PS
ZE
NS
PB
PB
PM
PS
ZE
NS
NM
PB
PM
PS
ZE
NS
NM
NB
PM
PS
ZE
NS
NM
NB
NB
PS
ZE
NS
NM
NB
NB
NB
ZE
NS
NM
NB
NB
NB
NB
Table 1: The control rules of the fuzzy controller of the shading (ΔuSH)
Δe
e
NB
NM
NS
ZE
PS
PM
PB
NB
NM
NS
ZE
PS
PM
PB
NB
NB
NB
NB
NM
NS
ZE
NB
NB
NB
NM
NS
ZE
PS
NB
NB
NM
NS
ZE
PS
PM
NB
NM
NS
ZE
PS
PM
PB
NM
NS
ZE
PS
PM
PB
PB
NS
ZE
PS
PM
PB
PB
PB
ZE
PS
PM
PB
PB
PB
PB
Table 2: The control rules of the fuzzy controller of the artificial lighting (ΔuAL)
4. Supervisor Architecture
The proposed control system can be referred to as intelligent control system
because the actions of the controller attempt to mimic high level decision
making processes of human operators. The architecture of supervisor unit is
shown in Figure 5 and the supervisor logic is presented in Table 3.
Supervisor
If Illuminance desired (r(k))  Natural Illuminance without shading (k)
Then
α1=0 → u*AL (k )  0
*
(k )  uSH (k )
α2=1 → uSH
Else
α1=1 → u *AL (k )  u AL (k )
*
(k )  0
α2=0 → uSH
end
Table 3: Supervisor (logic-based switching)
Illuminance
desired (r)
ery
e
Calculation of indoor natural
illuminance without shading
(Equation 1)
Supervisor
PI –FLC
(AL)
u AL
a1

u *AL
Zone illuminance (y)
ery
e
a2
PI –FLC
(SH)
uSH

TRNSYS
*
uSH
Figure 5: The architecture of proposed control system
5. Simulation Results
The performance of the two controllers is summarized in Table 2. The
performance criteria are the response performance, the illuminance
discomfort index, the natural lighting exploitation and the energy consumption
for electric lighting. The energy consumption is calculated for the one day
simulation period. In Figures 6 and 7 give the response performance of indoor
illuminance with and without complete exploitation of natural lighting. The
response of system output successfully approaches the set points. In the case
1 the control system involves important energy saving about 90% concerning
case 2. However, in the case 2 the control system does not achieve a low IDI
since the daylighting is one of the main reason’s that cause glare and visual
discomfort in occupants.
Figure 6: Response performance of
Indoor illuminance without the complete
exploitation of natural lightting (15η April).
Figure 7: Response performance of
Indoor illuminance under the complete
exploitation of natural lightting (15η April).
Performance of the zone level controllers (Illuminace tolerance=50lux)
Performance without the complete
Performance under the complete
exploitation of natural lighting (Case 1) exploitation of natural lighting (Case 2)
IDI=21.865 lux
IDI=27.570 lux
Natural lighting exploitation=56%
Natural lighting exploitation =94%
Response Performance
Overshooting: approximately zero
Steady state error: approximately zero
Response Performance
Overshooting: about 32%
Steady state error: approximately zero
Energy consumption (KWh/m2)
Energy consumption (KWh/m2)
Artificial lighting=31  10-3
Motor for shading=1,266  10-6
Artificial lighting=2,9  10-3
Motor for shading=2,055  10-6
Table 4: The performance without and with without the complete exploitation
of natural lighting
6. Conclusions
In this paper, we develop a supervisor of digital PI-like fuzzy logic controllers
for indoor lighting conditions control in buildings. The simulation results show
that the proposed fuzzy control system is achieved illuminance comfort and
important energy savings. The supervisory control system achieves energy
saving based on the full exploitation of the daylight.
Acknowledgements
The project is co-funded by the European Social Fund & National Resources EPEAEK II – ARCHIMIDIS.
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10
[11]
[12]
[13]
[14]
Argiriou, S. Lykoudis, S. Kontoyiannidis, C. A. Balaras, D. Asimakopoulos, M. Petrakis
and P. Kassomenos, “Comparison of Methodologies for TMY Generation Using 20
Years Data for Athens, Greece”, Solar Energy, vol. 66, no. 1, pp. 33-45, 1999.
S. Aydinli and J. Krochmann, “Data on daylight and solar radiation: Guide on Daylight”,
Draft for CIE TC 4.2, 1983.
ASHRAE, Handbook – Fundamentals, 2005.
A. I. Dounis, M. J. Santamouris, C. C. Lefas, “Building Visual Comfort Control with
Fuzzy Reasoning”, Journal of Intelligent and Fuzzy Systems, vol. 34, no. 1, pp.17–28,
1993.
A. Guillemin and N. Morel, “An innovative lighting controller integrated in a selfadaptive building control system”, Energy and Buildings, vol. 33, no. 5, pp. 477–87,
2001.
A. Guillemin, S. Molteni, “An energy-efficient controller for shading devices selfadapting to the user wishes”, Building and Environment, vol. 37, pp.1091–1097,
2002.
D. Kolokotsa, “Comparison of the performance of fuzzy controllers for the management
of the indoor environment”, Building and Environment, vol. 38, no. 12, pp. 1439–
1450, 2003.
C. P. Kurian, S. Kuriachan, J. Bhat, R. S. Aithal, “An adaptive neuro-fuzzy model for
the prediction and control of light in integrated lighting schemes”, Lighting Res.
Technoogy,. vol 37, no. 4, pp. 343-352, 2005.
Lah MT, Borut Z, Krainer A., “Fuzzy control for the illumination and temperature
comfort in a test chamber”, Building and Environment, vol. 40, pp. 1626–1637. 2005.
Lah MT, Borut Z, Peternelj J, Krainer A. Daylight illuminance control with fuzzy logic.
Solar Energy 2006;80:307-321.
DHW Li and JC Lam, “Measurements of solar radiation and illuminance on vertical
surfaces and daylighting implications”, Renewable Energy, vol. 20, pp. 389-404,
2000.
P. Littlefair, S. Ashton and H. Porter, "Luminous efficacy algorithms”, Joule 1 Program
– Dynamic characteristics of daylight data and daylighting design in Buildings, Final
Report, CEC Brussels, 1993.
M. Perraudeau, “Estimation of illuminances from solar radiation data”, Joule 2
DAYLIGHT II Program: Availability of Daylight – Design of a European Daylighting
Atlas, CSTB Nantes, 1994.
TRNSYS 16: A Transient System Simulation Program, Users manual, Solar Energy
Laboratory, University of Wisconsin-Madison, (2006).