Escolar Documentos
Profissional Documentos
Cultura Documentos
net/publication/283759053
Article in LEUKOS The Journal of the Illuminating Engineering Society of North America · November 2015
DOI: 10.1080/15502724.2015.1080122
CITATIONS READS
24 4,929
1 author:
Adam Sȩdziwy
AGH University of Science and Technology in Kraków
58 PUBLICATIONS 229 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Adam Sȩdziwy on 24 November 2015.
Abstract
This article compares two methods of roadway lighting design for situations
compliant with EN 13201-2 ME/MEW series. The goal is to find out which
one yields a more energy-efficient installation. The first method (further
referred to as the “default” one), used by industry-standard software, assumes
equally-spaced luminaires (in any arrangement) and uniform width of the
corresponding roadway. The second one (referred to as the “custom” one)
operates on exact inventory data including spacing, road width, mounting
height and other parameters given per luminaire rather than set up globally.
Both approaches are assumed to find solutions (installation setups) giving
minimal energy consumption which is accomplished, among others, by lamps
dimming. The article presents the performed computations, discusses the
obtained results and analyses the factors which might influence them. Results
show that using the custom computation method and actual inventory data
may bring an energy usage reduction at the level of 10% and more, compared
to the default approach.
Another issue critical from the practical point of view is the complexity of
the solution finding process. Although commonly-known programs allow for
building custom (non-uniform) scenes, the trial-and-error method of solution
finding, remarkably increases the design preparation time. Special heuristics
embedded in the custom method allow avoiding the complexity issue and en-
able bulk processing, which is crucial in the case of large-scale computations
made, for example, for large retrofits.
Keywords: street lighting, EN 13201, photometric computations, energy
efficiency, PhoCa
2
All common known lighting design software available on the market (DI-
ALux, Ulysse, Relux, AGi32 and others) by default uses the standard lighting
situations for road photometric computations (although it is also possible to
define custom scenes, e.g., using an AutoCAD underlay). In particular, both
road width (w) and luminaire spacing (s) are assumed to be uniform along
the entire road. Our experience shows that to ensure compliance with the
standards-derived requirements [CEN 2003a] lighting engineers set w and
s as the maxima of corresponding variables. Such an approach simplifies
computations and guarantees compliance with standards, but causes over-
illumination, and thus a cost overhead, as a side effect.
In this article we investigate the different approach to lighting installation
design, based on precise computations operating on actual data (luminaire-
by-luminaire) rather than on uniform lighting situations. Such computations
produce a precisely tailored design, which reduces power usage by 10% and
more. It should be stressed that the proposed method is fully compliant
with the standard regulating the calculation methodology and the obtained
results meet the requirements for a given lighting class. In the paper we show
that such an approach enables preparing lighting installations which require
less power compared to design methods used by default in industry-standard
photometric software.
The paper is organized as follows. In the next section the brief overview
of current design methods and solutions is presented. In Section 3 we discuss
in details precise photometric computations which produce a design for a
power-efficient lighting installation. The section also includes the analysis of
the method’s correctness, verification of results and the assessment of a com-
putational complexity. In Section 4 case studies are presented together with
quantitative results and in Section 5, some details regarding the performed
computations are discussed. Final conclusions may be found in Section 6.
3
Also, optimizations of power usage for LED-based installations has been in-
vestigated [Ylinen et al. 2011].
In the area of lighting design, customized photometric computations are
being developed. In the papers [Gómez-Lorente et al. 2013; Rabaza et al.
2013], computation methods based on genetic algorithms are presented. The
authors focused on optimizing uniformity and energy efficiency of lighting
installations. Although the GA-based method gives satisfying results, it is
intended for installations with evenly-spaced luminaires and a constant road
width rather than for real-life inventory data.
Existing lighting design software (e.g., DIALux [DIAL GmbH 2015], Ulysse
also known as TURBO Light [Urbis Lighting 2015], AGi32 [Lighting Ana-
lysts 2015]) enable finding energy-efficient solutions, but they lack scalability
and bulk processing capability, especially for non-uniform lighting situations
(e.g., non-evenly distributed poles). Their default calculation schemes rely
on standard road layouts containing evenly-spaced poles and constant width
carriageways.
3. Photometric computations
In this article we deal with optimizing power efficiency of a street lighting
installation by improving the photometric design process. Unlike the papers
mentioned in the previous section, it doesn’t introduce any novel computa-
tional methodology, but bases on the EN 13201-3 [CEN 2003b] standard,
which allows photometric computations on distorted road layouts differing
from uniform configurations for which this standard is defined. The problem
arising around photometric computations is the computational complexity of
the finding optimal configuration of lighting installation (in terms of energy
efficiency).
To illustrate this issue let us suppose that a retrofit of n = 200 road lu-
minaires is made. It covers fixture, pole and arm replacement. The following
variables are altered to find optimal solution: pole height: 8 m − 12 m with
0.5 m step (9 variants), arm length: 0.5 m − 3 m with 0.5 m step (6 variants),
inclination: 0◦ − 15◦ with 5◦ step (4 variants), dimming level: 0% − 50% with
1% step (51 variants; we assume that dimmings above 50% may be rejected
a priori), fixture model: pool of 500 models/photometric solids (500 vari-
ants). The total number of scenes to be calculated is N = 5, 508, 000 when
a uniform lighting situation is assumed or N = 5, 508, 000 × 200 ≈ 1.1 × 109
4
when deviations from uniformity are admissible (each luminaire needs to be
considered separately).
Such a complexity may lead to problems with practical usage of the ap-
proach unless appropriate computation methods are applied. The software
[Kotulski et al. 2013] being applied for those calculations uses heuristic tech-
niques resolving complexity issue and enabling bulk optimization (details
remain beyond the scope of this article).
3.1. Preliminaries
Lighting standard
In the paper we follow the European lighting standard, EN 13201-3, which
precisely defines the methodology of photometric computations. Although
this specification allows computations on irregularly-shaped areas, it uses
rectangular computation fields with longitudinal and transversal sizes which
are taken as the maximum or average luminaire spacing/road width measured
along a relevant carriageway. Such an approach ensures compliance with
performance requirements (EN 13201-2, see [CEN 2003a]) by assuming the
worst spatial conditions but may produce, as a side-effect, over-illumination
of particular road sections and thus increase energy-related costs. To resolve
this issue we propose an approach based on accurate computations, also
carried out according to EN 13201-3, which are performed field-by-field and
use the actual (not aggregated) inventory data.
It is assumed that both luminaire spacing and carriageway width corre-
sponding to an actual pole (see Fig.1) are non-constant parameters, varying
from luminaire to luminaire. This set may be easily extended over other
parameters, e.g., fixture inclination, pole setback and so on.
Computations presented below will be carried out for the ME series of
road lighting classes (EN 13201-2). In this standard, three groups of photo-
metric quantities are considered:
1. Luminance of the road surface of the carriageway for the dry road sur-
face condition, which is described by the average road surface luminance
(L̄), the overall uniformity of the luminance (Uo ) and the longitudinal
uniformity of the luminance (Ul ).
2. Disability glare: the threshold increment (T I).
3. Lighting of surroundings: the surround ratio (SR).
5
Energy efficiency assessment
To assess and compare energy efficiency of two solutions, i.e., based on
the aggregating (i.e., default) and exact approach respectively, we calculate
the total power usage, Ptot , assuming that a LED fixture total luminous flux
is linearly dependent on the power supplied, i.e.,
Pi Φi
= ≡ LF Ri , (1)
Pnom,i Φmax,i
where Pnom,i is the nominal power of the i-th fixture, Φmax,i is the total
luminous flux reached for Pnom,i supplied, Φi denotes the total luminous flux
for Pi and LF Ri (referred to as the luminous flux ratio) is the corresponding
dimming coefficient such that LF Ri = 0 and LF Ri = 1 represent fully
dimmed and non-dimmed states, respectively. Making such assumption we
get:
Xn Xn
Ptot = Pi = LF Ri Pnom,i . (2)
i=1 i=1
The relative power reduction being the quantitative measure of the solution
quality is calculated as:
Pn
Pnom − Ptot (1 − LF Ri )Pnom,i
Psav = = i=1 Pn . (3)
Pnom i=1 Pnom,i
Note that Psav ∈ [0, 1]. For two solutions the better will be the one with a
higher Psav .
6
w1 w2 w3 w4
(a) s1 s2 s3
l1 l2 l3 l4
w1 w2 w3 w4
(b) s1 s2 s3
l1 l2 l3 l4
Figure 1: Two lighting situations considered in computations with sample computation
fields (gray shaded): (a) single sided right, (b) staggered right. Luminaires (l1 , l2 , . . . )
are marked as bold dots. Luminaires located at the left side are numbered with primes:
10 , 20 , . . . .
Inventory
Let I = (l1 , l2 , . . . ln ) be an inventory storing the actual data of n ordered
luminaires where a luminaire record, li , contains the following fields:
• ID – luminaire identifier,
• sb – pole setback,
• oh – fixture overhang,
7
• M F – maintenance factor,
The value of field x for a luminaire record li will be denoted as [x]i . The
mentioned ordering of luminaires is set straightforwardly as their location
order along the road.
The next assumption is related to luminaire spacing,
and road width, [width]i (i ≤ n), which are crucial for the photometric calcu-
lations. In the trivial case, si and [width]i are assumed to be constant. The
effect of deviations from even spacing was examined by looking at normal,
double overlapped normal, and uniform distributions. The latter two distri-
butions are described in detail in Section 5.3. Our initial in-depth analysis
was based on normal distributions:
8
3.2. Computations
The computation method presented below works for any type of the lu-
minaire arrangement, along either a single or a double carriageway.
The general scheme of computations is following. First, six initial lumi-
naires are taken to cover an area taken as a “standard” lighting situation in
the default computation approach. For those luminaires the default com-
putation method is used. Such a number of initial poles is assumed with
reference to the standard [CEN 2003b], which specifies that the boundary
furthest from the observer should be at least 12 × H (where H is the mount-
ing height) from the calculation point in the direction away from the observer
and the boundary nearest to the observer should be at least 5 × H from the
calculation point in the direction towards the observer. For mounting height
H = 12 m and luminaire spacing s = 35 m we obtain the number of lumi-
naires of N = 17 × H/s ≈ 6 required to proceed with the standard compu-
tation method. It should be noted that the influence of using the standard
method for determining LF Rs of initial luminaires may be neglected from
the perspective of resultant power savings, especially when the total num-
ber of luminaires is significantly greater than N . Moreover, the impact of
overlighting generated in this phase, i.e., setting up overestimated LF Rs for
six initial fixtures, is compensated by lowered LF R values for the few sub-
sequent ones, depending on the spacing value. Lighting class requirements,
however, remain satisfied.
After completing calculations for initial fixtures one proceeds with sub-
sequent luminaires using the exact method until the last pole is reached (see
Algorithm 1).
In the first, “start-up” phase (lines 2-4) of the optimization process, six
initial luminaires are taken and an iterative search of an optimal value of
LF R is made. This phase complies with the “default” EN 13201-3 calculation
scheme.
Next, the algorithm proceeds luminaire after luminaire (line 5), with ac-
curate calculations, determining [LF R]i for subsequent fixtures. This part
of the algorithm is supported by the PhoCa program, the custom computa-
tion tool further referred to as the custom tool, which enables time-efficient
multivariate optimization of lighting installation performance [Kotulski et al.
2013]. It uses heuristic methods for efficient search performed in lines 4 and
10. It should be remarked that the search space is restricted to the LF R
9
Algorithm 1: CustomOptimization
Input : I = (l1 , l2 , . . . ln ) – luminaire inventory representing the
installation,
C – lighting class
Output: LF Rs of subsequent luminaires, minimizing the
installation’s power usage, ensuring the compliance with the
class C requirements
1 begin
2 S ←Six initial luminaires from the inventory;
3 (W, L) ←(maximum width in S, maximum spacing in S);
4 LF R ←minimal value of luminous flux ratio common for S, such
that lighting performance computed for luminaire spacing L and
road width W meets requirements of the class C;
5 foreach subsequent i-th luminaire, li , following those from S,
except the last one do
6 W ←select greater road width from {[width]i , [width]i+1 };
7 L ← spacing for i-th luminaire;
8 Create computational field F of the size W × L;
9 lf r ←the minimum value of luminous flux ratio satisfying C
class requirements on F ;
10 For li and all luminaires following li : LF R ← lf r;
11 For the last luminaire, ln , in the inventory: LF R ←select the highest
LF R of three preceding luminaires;
10
variable only in the presented algorithm. In the general case, however, it
may be a cross product corresponding to several variables such as LF R, arm
length, spacing, fixture model, fixture’s inclination, mounting height and so
on. In such a case, computations may be not feasible in a reasonable time
when using naive, brute force methods. Heuristics used by the custom tool
allow avoiding those complexity-related problems.
When computing luminance, illuminance, uniformities and so forth, one
has to take into account a set of fixtures affecting a computation field [CEN
2003b]. In the presented approach their locations and other properties are not
aggregated (e.g., averaged) but taken as is, using [(x, y)]i , [H]i , [sb]i , [oh]i and
other fields. A luminaire li (starting with i = 6) is taken and a computation
field Fi located between li and li+1 is analyzed. For these cases we set the
width and length of Fi : W = max{[width]i , [width]i+1 } and L = si . Note
that [LF R]k is already known for k ≤ i, as they were determined in previous
steps. Thus, we modify [LF R]k for k > i only.
Since the adjustments of subsequent fixtures change in the course the
computations and thus may influence photometric conditions of previously
proceeded fields Fi , the important step after completing Algorithm 1 is verify-
ing if the resultant settings (in our case: [LF R]i values) make the installation
performance meet the lighting class requirements. It is achieved by check-
ing the compliance with a given standard along the entire installation: one
has to perform photometric computations for each field Fi and verify if the
obtained values are acceptable. This verification will be referred to as an
overall compliance test.
11
The second step of verification is based on the observation that the accu-
rate method should be convergent to the “default” approach based on uniform
luminaire spacing and constant road width. In the other words, the accurate
computations have to produce the same results as the “default” method, when
the deviations of w and s tend to zero. In particular, the results obtained for
an uniform layout of luminaire arrangement (lines 2-4) has to be the same as
for the next computation fields proceeded in subsequent iterations (see the
loop in the line 5), when standard deviations are small enough.
Let us denote the set of photometric parameters (for testing of compliance
with the ME series) computed using the “default” method as
From the practical point of view, it is sufficient to verify the fulfillment of the
equality 7 by putting σw = σs = 10−3 m. Performed verification of the method
correctness proved full convergence of the proposed accurate approach to the
standard computational method.
12
4. Case study
While investigating the impact of accurate computations on the resultant
power usage of an installation we carried out two types of tests. In the first
group of tests we analyzed power reductions obtained thanks to applying the
accurate approach, for particular lighting classes of the ME series. In those
cases the arbitrary road parameters, lighting installation properties, means
and standard deviations for normal distributions were used.
For the tests of the second type we selected one lighting class and in-
vestigated how the geometric properties of the road and lighting installation
(in terms of carriageway width and luminaire spacing distributions, see 5)
influence the power reductions being computed.
For both test groups we used fixed values of the following parameters:
M F = 0.85, oh = 0.0 m, sb = 0.0 m (which means that a fixture nadir
is located exactly over a carriageway border), azm = 0◦ . For the second
group we additionally fixed inc = 0◦ . The road surface was described by the
R-Table R3 with Q0 = 0.07; observer age was set to 23 years.
Table 1 presents values of remaining parameters used in computations
and the obtained results of tests of the first type:
1. Luminaire arrangement which was either staggered right or single
sided right.
2. Road width which was assumed to be variant and defined by the nor-
mal distribution with both the average and standard deviation given.
3. Luminaire spacing, similarly as the road width, defined by the nor-
mal distribution with the average and standard deviation given.
4. Fixture – the name of an LED fixture model which photometric solid
was used in computations (fixtures from three different vendors were
used). For each fixture the nominal wattage and luminous flux are also
presented. Photometric solids were read from profile files provided by
particular manufacturers.
5. Number of lanes of a carriageway, fixture mounting height and
inclination used in a given case.
6. Computation field size in the “default” approach denotes dimensions
of the form W × L, where
W = max [width]i , L = max si .
i=1,2,...n i=1,2,...n−1
Note that neither road width nor luminaire spacing are assumed to be
uniform. For that reason maxima for both are selected.
13
7. LFR (for the “default” approach) is the fraction of the nominal fixture
power which has to be supplied to meet the lighting class requirements,
as calculated using the “default” approach (see point 6).
8. The Power savings field represents the ratio defined in 3, where Ptot
was introduced in 2 and Pnom is replaced by the power usage corre-
sponding to the LF R given in the previous point. In the other words,
it expresses the relative power reduction when migrating from an LF R
obtained in the “default” manner to LF Rs determined in accurate com-
putations.
The set of luminaires used for computations consisted of N = 100 luminaires
sampled with distributions defined in point 3. Corresponding road widths
were sampled with distributions defined in point 2.
Remark. It may be seen that for some cases (especially for ME5 and ME6
lighting classes) we selected fixtures with oversized wattages; this is reflected
in corresponding LF R values (50% and 45% for ME5 and ME6, respectively,
for the “default” approach). Although it may be an issue from the business
point of view (fixture price) we neglected it here as not influencing technical
conclusions.
Table 2 contains the comparison of the results obtained for all test cases
(Table 1) in the “default” approach, using the custom tool and Ulysse respec-
tively: only the few neglectable discrepancies are present.
For the tests of the second type we selected the ME4a lighting class
and assumed the following parameters of the road and installation. The
GE Odyssey LED 71W 3000K fixture with 0◦ inclination and 10 m mounting
height was applied. The average road width and luminaire spacing were set to
wavg = 7.5 m and savg = 25 m, respectively. The set of randomly generated
luminaires contained N = 50 luminaires which covered an approximately
1250 m long road section. Standard deviations of w and s were sampled
with step ∆ = 0.03 m in the ranges [10−5 m, 0.75 m] for the road width and
(i) (j)
[10−5 m, 3.5 m] for the spacing. For each (σw , σs ) sample the resultant Psav
was computed. Figure 2 shows the results of those computations. It should
be remarked that for too high standard deviations it may be impossible to
meet lighting class requirements unless either the maximum luminous flux
or the photometric solid is changed. For each completed computation cycle,
the overall compliance test was successfully performed (see Section 3.2).
14
Parameter ME1 ME2 ME3a ME4a ME5 ME6
Staggered Staggered Single Single Single Single
Luminaire arrangement
right right sided right sided right sided right sided right
Avg [m] 11 11 8 7.5 7.5 7.5
Road width
Std dev 0.2 0.2 0.3 0.3 0.2 0.2
[m]
Avg [m] 30 30 30 25 18 18
Luminaire spacing
Std dev 2 2 2 2 1 1
[m]
Evolve
Ampera
Rdwy
Mini: Flat
Model BGP353 T25 A Scalable 3 Odyssey LED 71W
Glass 5137
Fixture ECO226-2S/657 Mod 3000K
16 XP-G2
Roadway
Wattage 207 172 71 36
15
[W ]
Lum. flux 22 635 13 800 5 273 3 603
[lm]
Number of lanes 3 3 2 2 2 2
Mounting height [m] 10 10 12 10 9.5 8
Inclination [deg] 5 5 0 0 10 10
Computation field 33.74×11.55 32.17×11.70 35.97×8.62 29.27×8.06 20.32×8.06 20.25×8.16
size in the “default”
approach [m2 ]
LFR (“default” approach) 98% 70% 90% 99% 50% 45%
Power 8.9% 5.7% 14.8% 13.1% 10.1% 11.9%
savings
Table 1: Test cases
ME1 ME2
Custom Ulysse Relative EN Custom Ulysse Relative EN
tool difference 13201-2 tool difference 13201-2
diff reference diff reference
value value
Lavg [ mcd2 ] 2.00 2.00 0% 2.00 Lavg [ mcd2 ] 1.50 1.49 1% 1.50
Uo 0.72 0.72 0% 0.40 Uo 0.72 0.72 0% 0.40
Ul 0.85 0.85 0% 0.70 Ul 0.85 0.85 0% 0.70
T I [%] 3 3 0% 10 T I [%] 2.2 2.7 0% 10
SR 0.55 0.55 0% 0.50 SR 0.55 0.55 0% 0.50
ME3a ME4a
Custom Ulysse Relative EN Custom Ulysse Relative EN
tool difference 13201-2 tool difference 13201-2
diff reference diff reference
value value
Lavg [ mcd2 ] 1.00 1.00 0% 1.00 Lavg [ mcd2 ] 0.75 0.75 0% 0.75
Uo 0.60 0.60 0% 0.40 Uo 0.45 0.45 0% 0.40
Ul 0.76 0.76 0% 0.70 Ul 0.78 0.78 0% 0.60
T I [%] 8.4 8.4 0% 15 T I [%] 4 4 0% 15
SR 0.88 0.88 0% 0.50 SR 0.51 0.51 0% 0.50
ME5 ME6
Custom Ulysse Relative EN Custom Ulysse Relative EN
tool difference 13201-2 tool difference 13201-2
diff reference diff reference
value value
Lavg [ mcd2 ] 0.51 0.50 2% 0.50 Lavg [ mcd2 ] 0.30 0.30 0% 0.30
Uo 0.61 0.61 0% 0.35 Uo 0.57 0.57 0% 0.35
Ul 0.88 0.88 0% 0.40 Ul 0.91 0.91 0% 0.40
T I [%] 3.9 3.9 0% 15 T I [%] 6.6 6.7 -1% 15
SR 0.54 0.54 0% 0.50 SR 0.49 0.50 2% N/A
Table 2: Comparison of results obtained using the custom tool and Ulysse. Relative
difference, diff, is calculated using 6.
16
Figure 2: 3D Plot representing power savings as a function of standard deviations of
spacing and carriageway width.
Remark. After computing [LF R]i the question may be posed regarding the
visual effect of those (potentially) non-uniform LFR values. Although this
issue is not critical from our point of view, it may be taken into account in
real-life situations. For the case considered above and σw = 0.38 m, σs =
1.75 m (midpoint in Fig.2 mesh), we obtain the LFR distribution presented in
Figure 3 with the average LF R = 0.41 and the standard deviation σLF R =
0.02. These results may be regarded as not influencing the overall visual
effect.
5. Discussion
In this section we discuss some questions concerning assumptions under-
lying tests presented above.
17
25
20
Frequency [%]
15
10
0.38
0.39
0.40
0.41
0.42
0.43
0.44
0.45
0.46
LFR
The conclusion of this simulation is that the mean is not a suitable ag-
gregating function in this context.
In the next subsection, we describe how results of average-based computa-
tions depend on spacing/width standard deviations, σs and σw , respectively.
18
60
Frequency [N ]
40
20
0.62
0.63
0.64
0.65
0.66
0.67
0.68
0.69
0.70
0.71
0.72
0.73
0.74
Lavg [cd/m2 ]
Figure 4: Lavg histogram for non-compliant scenes. Lavg = 0.75 cd/m2 is the minimum
acceptable value for ME4a class.
19
Figure 5: The ratio of calculation fields not complying to ME4a as the function of standard
deviations of luminaire spacing and road width
constant. The meaning of both tests is similar. They inform what the
averaged value of non-compliant Lavg is for different values of σs and
σw . Such information may be useful for cases when the value of Lavg
being slightly below the standard-defined level is admitted.
20
Figure 6: The ratios of calculation fields not complying to ME4a and corresponding aver-
aged values of Lavg
1 (column ME4a). Other parameters (inclination, setback and so on) are also
the same.
1. In this case, distributions being sums of two normal ones are taken:
√ √ 2 2
fS (s) = π(1 + 2) e−(s−20) + e−0.5(s−24) ,
√ 2
fW (w) = 2π e−2(w−6.5) + e−2(w−7.5) ,
6. Conclusions
In this article we proposed photometric computations compliant with EN
13201-3, addressing irregular lighting situations, i.e., when neither luminaire
spacing nor road width are constant. Such an approach to real data pro-
cessing differs from the typical usage of the EN 13201-3 method where both
21
Figure 7: Histograms of luminaire spacing (upper) and road width (lower) generated using
respectively fS (s) and fW (w) distributions.
spacing and road width are taken as averages and/or maxima. Using actual
data instead of aggregated ones yields a significant reduction of the resultant
power consumption: for considered cases it reached nearly 15%. Tests showed
that this reduction grows with increasing standard deviations of luminaire
spacing and road width. The obtained results are notably better than those
produced by industry-standard software in terms of energy efficiency of the
lighting installation being designed. Computations presented in the article
were performed using the custom software developed at AGH University of
Science and Technology.
The custom computations-based approach using this software was applied
within the ISE project sponsored by Polish National Fund for Environmen-
tal Protection and Water Management [ISE 2014]. The project aimed at
retrofitting the street lighting in the city of Cracow and covered 3800 LED
fixtures.
Acknowledgment
The work is part of a research project supported by the European Union,
Human Capital Operational Programme, SPIN project no. 502.120.2066/C96
by ISI Center.
22
References
[CEN 2003a] European Committee For Standardization. 2003a. CEN 13201-
2. Road lighting - part 2: Performance requirements. Brussels (Belgium):
European Committee for Standardization. 16p.
[CEN 2003b] European Committee For Standardization. 2003b. CEN 13201-
3. Road lighting - part 3: Calculation of performance. Brussels (Belgium):
European Committee for Standardization. 41p.
Dazzletek. 2015. Regulus. <http://goo.gl/ZcJAvL> Accessed 2015 June 9.
DIAL GmbH. 2015. DIALux. <http://goo.gl/21XOBx> Accessed 2015 June
9.
Echelon. 2015. Echelon. <http://goo.gl/mfpGYT> Accessed 2015 June 9.
GE Lighting. 2015. LightGrid. <http://goo.gl/cZSBY3> Accessed 2015
June 9.
Gómez-Lorente D, Rabaza O, Espín Estrella A, Pena-García A. 2013. A
new methodology for calculating roadway lighting design based on a multi-
objective evolutionary algorithm. Expert Systems with Applications 40(6):
2156 – 2164.
Kostic M, and Djokic L. 2009. Recommendations for energy efficient and
visually acceptable street lighting. Energy 34(10):1565 – 1572.
Kotulski L, De Landtsheer J, Penninck S, Sędziwy A, Wojnicki I. 2013. Sup-
porting energy efficiency optimization in lighting design process. Proc. of
12th European Lighting Conference Lux Europa, Krakow, 17-19 September
2013. <http://goo.gl/tfbPf3> Accessed 2015 June 9.
Lighting Analysts. 2015. AGi32. <http://www.agi32.com> Accessed 2015
June 9.
Philips. 2015. CityTouch. <http://goo.gl/GhCJsX> Accessed 2015 June 9.
23
Rabaza O, Pena-García A, Pérez-Ocón F, and Gómez-Lorente D. 2013. A
simple method for designing efficient public lighting, based on new parameter
relationships. Expert Systems with Applications, 40(18):7305 – 7315.
Schréder. 2015. Owlet. <http://goo.gl/9g53OP> Accessed 2015 June 9.
Urbis Lighting. 2015. TURBO Light. <http://goo.gl/FY8yUX> Accessed
2015 June 9.
Ylinen AM, Tähkämö L, Puolakka M, Halonen L. 2011. Road lighting qual-
ity, energy efficiency, and mesopic design-LED street lighting case study.
Leukos, 8(1):9–24.
24