Supporting energy efficiency optimization in
lighting design process
Leszek Kotulski1, Jeroen De Landtsheer2, Sven Penninck2, Adam Sędziwy1, Igor Wojnicki1
1
Department of Applied Computer Science, AGH University of Science and Technology,
Al. Mickiewicza 30, 30-059 Kraków, Poland,
2
EANDIS, Gebouw Waterkant, Brusselsesteenweg 199, 9090 Melle, Belgium
{kotulski,sedziwy,wojnicki}@agh.edu.pl,
{jeroen.delandtsheer,sven.penninck}@eandis.be
Abstract. In the article we discuss the properties of a designer-friendly solution finder
based on accurate photometric computations. Such software automates reconfigurations
and recalculations of a scene implied by changes introduced to a project by designer,
keeping a balance between visualization features and availability of quantitative
photometric data. The goal of the redesign in achieving power usage reduction. In the
article we also deal with the practical approach to the problem presented by the Eandis,
the Belgian distribution system operator.
1. Introduction
The existing software market supplies a range of tools supporting the lighting design
process. Basically it offers photometric calculations with limited design capabilities. In
particular, most of them are limited to street-like configurations and none allow for design
optimization toward energy efficient solutions. The new developed software will allow
optimizing lighting on different public areas (streets, sidewalks, squares, parks) and will
take into account smart control including dimming, presence detection and so on .
In the article we discuss the issue of filling this gap by preparing a designer-friendly
solution finder, based on accurate photometric computations. Such software automates
reconfigurations and recalculations of a scene implied by changes introduced to a project
by designer, keeping a balance between visualization features and availability of
quantitative photometric data.
The software is tested with practical cases provided by a grid operator, to determine
energy savings for cities and communities.
This project is the effect of collaboration between AGH University of Science of
Technology in Cracow and Eandis (the Belgian grid operator for electricity and gas). AGH
University of Science and Technology developed software and Eandis provided AGH UST
with test cases and did testing on the developed software.
The article is organized as follows: Section 2 contains the overview of lighting design
process paradigm and in Section 3 we consider the workflow in a typical lighting design
task. Optimization of the street lighting energy efficiency and its computational
complexity are presented in Section 4. The impact of the control for energy efficiency is
discussed in Section 5. In Section 6 photometric calculation software, PhoCa, created in
AGH UST, is presented. Section 7 focuses on the practical approach to street lighting
1
optimization on the example of Eandis grid operator practice. The summary and the
directions for future works are included in Section 8.
2. Outdoor Lighting Design Paradigm
Outdoor lighting design process may be considered from several perspectives,
dependently on a role played by its participants [4]. Objectives of an architect are often
not fully compliant with a street light electrician ones, and require the cooperation of both
parties. Yet another perspective may be represented by an investor or a customer. This
multiplicity of approaches makes the design be a complex and iterative process with
multiple interactions among particular actors. Obviously, this enlarges a design time.
The paradigm of the outdoor lighting design relies on three basic elements described
below:
Aesthetics – this issue plays a special role in architectural lighting, e.g., in highlighting
buildings, monuments, pavements, squares and so on. Objectives related to the aesthetics
have to be agreed between an architect and an investor. Sometimes they are constrained
by some local regulations, obligatory for new investments.
Functionality includes both the compliance with existing lighting standards (e.g., EN
13201:2) and the support for social goals such as public safety [3], improved orientation,
highlighting identity of places and others [1].
Cost-effectiveness is the only fully measurable criterion of the lighting design assessment.
The cost-effectiveness is a resultant of several factors:
1. Investment costs
2. Maintenance costs
3. Energy prices
4. Fixture energy efficiency
Let us note that cost-effectiveness may be expressed in various ways, for example in
terms or investment payback period, net present value (NPV) or, in the most intuitive
case, in terms of the energy usage reduction. It should be strongly emphasized, however,
that fulfilling the last postulate does not guarantee optimizing a cost-effectiveness, which
depends on other parameters, as it was mentioned above. Moreover it is possible that for a
given power level we have several alternative solutions [2].
One may distinguish three basic types of actors involved in various phases of a design
process (Fig.1). Functionality of lighting design software has to satisfy their needs.
Figure 1. Three actors of the design process.
2
Designer perspective
The first actor is a designer. The required software capability as seen from this perspective
is suggesting an optimal configuration of a lighting solution to be prepared. It should be
remarked that the understanding of an optimality notion varies dependently on primary
objectives accepted for a given design. For example, when we consider a retrofit of street
lighting and if the power usage reduction is an only issue then software should suggest
fixture models (selected from a defined pool of fixtures) which imply the minimum energy
consumption and ensure the accordance with corresponding lighting standards.
Investor perspective
An investor is the second actor of a lighting design process. As an entity capable of
investment decisions making he needs to possess all information necessary to do it. In this
context, software is expected to provide a comprehensive and detailed information about
the structure of potential gains obtained thanks to selecting a given re-design scenario
(e.g., which areas of a lighting solution yield the highest power reduction). Also suggesting
other, alternative scenarios accompanied with corresponding pros and cons, is very
helpful. The additional software function, necessary for decision making is generating
detailed quantitative reports describing how a new design compares to an existing one.
Customer perspective
The third actor, which is referred to as a customer, may be partially identified with
investor. There doesn’t exist, however, a straightforward relation between both roles. The
software capability considered for this actor is (apart from finding the solution being
energy-optimal, payback-optimal and so on) reduction of the design life cycle time. To
accomplish that a program should provide multi-scenario output. Thus several iterations of
a project development, made between a customer and a designer, may be reduced to a
single one. Regarding the single solution design time, this reduces significantly overall
design time.
3. Design Process
The lighting design process (see Figure 2) is started by a lighting designer or architect.
Spatial and compositional assumptions regarding the architectural space are made
resulting in conceptual sketch. The design at this stage is very informal, sketching software
is being used (e.g. Google SketchUp). The sketch represents a general view of the scene
with light points indicated and their general parameters in terms of light cones. It is manly
to identify where the light points should be and where the actual light should go.
Figure 2. Design process.
3
Then the sketch is transformed into a two or three dimensional (2D, 3D) technical drawing
resulting in a spatial concept, a wire-frame. This allows to precisely specify places which
are to be illuminated, color, temperature, quality and other parameters of a lighting
composition. The drawing is performed by a supporting software such as: AutoCad,
ArchiCad, Revit etc.
A next step, verification, is performed by a lighting engineer. Luminaires and
intensities of light sources are selected, according to the assumptions provided by the
designer. Furthermore, they are verified using photometric software (Dialux, Calculus,
Ulysse or similar) if technical capabilities of luminaries meet requirements of the project.
This phase impacts number, power and detailed specifications of the luminaires. Since
the technical drawing (wire-frame) can contain multiple elements not influencing
photometrics, it has to be tuned accordingly or even created from very beginning.
Next, the parameters calculated in the previous stage are given to the lighting
designer to verify. It can be achieved through preparing a three dimensional, photorealistic visualization, a 3D model. It is supported by yet another software (e.g. 3ds Max,
Maya). A final effect is analyzed. If it does not satisfy the designer it is adjusted
accordingly and the process loops back to the spatial concept (wire-frame) or verification
(Photometric calculations) stages.
Based on the adjustments of the 3D model the results from previous steps need to be
updated. These steps are performed iteratively until satisfying results are achieved, being
a trial and error process. Since each stage is isolated some errors or artifacts can be
introduced unwillingly in the process. This leads to inconsistencies and lengthens the
entire process.
It needs to be pointed out that there is lack of automation between subsequent
stages. Data produced as a result of the conceptual stage need to be interpreted and
recreated as a wire-frame model and so on. Human interactions are required for transiting
data between subsequent stages.
The most problems are caused by taking the corrections and feeding them back to the
verification and adjustments (photometrics) stage. Multiple iterations, to achieve a
satisfying result, might cause even more mistakes and elongate the entire process.
In general entire process can be optimized by: reducing number of human oriented
interactions and increasing effectiveness of selecting and applying parameters to light
points (see Figure 3).
Reducing number of human-oriented interactions is to unify data interfaces among
applications to ensure proper import and export and their automation. Data flow among
applications should be provided with minimal human interactions. Calculation and
application light point parameters to the design, including corrections, contribute
significantly to the effectiveness of the entire process. In this stage actual optimization
criteria are applied which may vary depending on requirements. Theses are: energy
consumption reduction, public safety increase, overexposure elimination etc. From this
perspective the design is perceived as yet another set of constraints in addition to the ones
mentioned above.
4
Figure 3. Desired design process.
The main focus regards the loop, which is transitions: 3, 4 and 5 (Figure 3). It covers
photometric calculations, optional 3D visualization, and applying corrections to the design,
which require recalculations in turn. The key element is to improve quality, precision and
usability of photometric calculations which have been identified as the bottleneck. A
prototype software providing the calculations and optimizing light points parameters have
been implemented. Automation of transitions is indicated accordingly (compare with
Figure 2).
4. Energy Efficiency in Design Process. Computational
Challenges
There are several common known, methods of reducing a lighting system energy usage
such as: (i) installing low wattage luminaries, e.g. LEDs, (ii) adjusting a system
performance to the actual state of an environment, (iii) reducing a road class (in terms of
EN 13201:2 requirements) for the night time, and adjusting the performance
characteristics accordingly, (iv) implementing the proper policy of the non-recoverable
light loss compensation. Their applicability varies dependently on an actual design
background (new installation, retrofit, fixture types to be used, financial constraints and
so forth).
The decline in LEDs prices together with their continuous technological development
make LED to be the most promising technology regarding the approaches listed above. On
the other side, however, multiple LED-based street lighting installations require using
proper software tools oriented for obtaining maximum benefits of such advanced
properties. Those properties (stepless dimming, extremely short onset times or possibility
of preparing user-defined photometric solids) enable preparing solutions profiled strictly
for a particular customer.
The closer look at that reveals the practical problem which influences a design task.
This problem is related to the complexity of finding a design solution meeting obligatory
photometric requirements. To illustrate above let us consider the following scenario of
designing a new street lighting for a road of a given class, say ME4a (according to DIN EN
13201:2, see Table 1). It is assumed that parameters of an installation, listed in the Table
2 may be varied in a process of finding an optimal solution. The table presents assumed
variability ranges, corresponding step sizes and resultant number of variants for each
parameter. The total number of variants, N, obtained from the Table 2 is
5
N=41 x 21 x 3000 x 21 x 41 x 51 = 1,13 x 1011.
Even if applying some heuristics based on expert knowledge/experience or good practices
(e.g., assuming that the optimum distance between two neighboring poles is determined
by the position of the ground point where the maximum candlepower outputs meet) the
search space still remains too big to find a solution in a reasonable time, in particular with
no computer support.
Table 1. ME lighting classes according to EN 13201:2
Class
Lw
[cd/m2]
Uo
UI
TI %
SR
ME1
2,0
0,4
0,7
10
0,5
ME2
1,5
0,4
0,7
10
0,5
ME3a
1,0
0,4
0,7
15
0,5
ME3b
1,0
0,4
0,6
15
0,5
ME3c
1,0
0,4
0,5
15
0,5
ME4a
0,75
0,4
0,6
15
0,5
ME4b
0,75
0,4
0,5
15
0,5
ME5
0,5
0,35
0,4
15
0,5
ME6
0,3
0,35
0,4
15
n/a
Due to the combinatorial explosion implied by the high number of degrees of freedom
in optimization tasks, the new approach to the photometric computations has to be
applied. As the brute force method (i.e., testing all potential solutions contained in a
search space) fails for problems similar (or more complex) to the above one, more
sophisticated methods based on artificial intelligence computational techniques are
applicable here.
6
Table 2. The exemplary list of variables for an optimization process
Parameter name
Start value
End value
Step
Number of variants
Lamp spacing [m]
20,0
40,0
0,5
41
Pole height [m]
8,0
12,0
0,2
21
Fixture type
n/a
n/a
n/a
30001
0
20
1
21
Overhang [m]
-2,0
2,0
0,1
41
Dimming [%]
0
50
1
51
Tilt [deg]
5. (Re)Design and Control
Economic profits (we focus on the power usage only, neglecting other factors like a
payback period) implied by using an advanced lighting technology, e.g, LED, may be
obtained twofold. Firstly, the luminous efficacy and lifetime of a light source (see Table 3)
bring power savings through the replacement only (retrofit). According to manufacturers’
data2, energy usage reduction can reach the level up to 40-45%. The mentioned properties
of LED lamps (such as stepless dimming or low onset time), however, create possibilities of
further savings achieved by implementing suitable control schemes. The concept of a
responsive lighting control system relies on cooperation of three elements: lighting
infrastructure (luminaries, power cabinets, power lines), telemetry layer (including various
types of sensors) and control center, coupling the first and the second.
Design of street lighting installations is usually considered separately from a lighting
control system. For lighting systems covering larger areas it is advisable to merge both
perspectives. In such a case lighting design gets augmented by adding the telemetry layer
but also by affecting objectives which have to reflect the overall system performance.
1
The number of 3000 is a rough approximation of the number of outdoor fixture models/types
produced by three leading manufacturers, assuming that each of them offers at least 1000
models/types. (The last assumption seems to be reasonable: General Electric LED fixture,
Cobrahead, offers up to 650 photometric solids). Obviosuly, some types will be a priori rejected as
not compatibile with a technical specification.
2
E.g., those supplied by GE Lighting.
7
Table 3. Light source characteristics3
Onset time†
Light Source
Efficacy (lm/W)
Life (hours)*
Tungsten filament lamp
12-20
750-4000
Fluorescent (incl. compact)
60-100
10,000-30,000
1-60 s
Metal halide
80-110
10,000-20,000
60-300 s
Xenon
30-60
1000-5000
1 μs
Light-emitting diode (white)
90-130
50,000-100,000
10-20 ns
0.1-0.3 s
*Assuming steady operation. †Time to reach 90% of maximum light output.
6. PhoCa Software
Within the international R&D project, Products and Services of a Living Smart Energy
City Lab, carried by AGH University of Science and Technology, EANDIS and other partners,
the prototype software, PhoCa (abbrev. of Photometric Calculations) is developed, for
optimizing street lighting installations (Figure 4).
Figure 4. PhoCa screenshot
Besides typical functionality of photometric software like Dialux, Ulysse, Calculux
Road and so on, program offers an optimization performed with respect to the userdefined criteria. Moreover, new algorithms reducing the computation time are applied.
Tests conducted with real data provided by EANDIS, showed that tuning of working
parameters of existing street lighting installations based on high pressure discharge lamps,
brings power usage reduction of the order 5-8%.
3
Source: John D. Bullough, Yiting Zhu, Nadarajah Narendran,Characteristics of Light-Emitting Diode
Sources: Relevance for Visual Signal Detection, Lighting Research Center, Rensselaer Polytechnic
Institute, 21 Union Street, Troy, NY 12180, USA, (2012)
8
Figure 5. 3D visualization of photometric calculations with Maya
7. Improving Energy Efficiency – Practical Approach
Public lighting consumes a lot of energy. The energy cost of public lighting takes in
general 40% to 50% of the city total energy budget . As it was mentioned earlier, new
energy savings (and reductions in energy costs) can be realized with investments in energyefficient public lighting installation. Thus the actual problem is: how to find the energy
efficiency potential on existing and new public lighting installations ?
Following input is needed to answer this question:
The required lighting class for each road (ME, CE, … classes; according to EN
13201:2).
(ii)
The detailed road configuration (road width, spacing, position of the luminary
according to the road, actual fixture, actual type of lamp, actual power,
mounting heights, etc.).
(iii)
Lighting calculations to compare the actual lighting installation with new
optimized (energy-efficient) one(s).
Today distribution system operators (abbrev. DSOs) are bumping into several issues
when getting this input. They will be highlighted below.
(i)
Different lighting classes for same type of roads-streets
EN 13201 standard offers the method of finding a lighting class for each road type. This is a
subjective method, depending on the interpretation of an expert and for that reason this
gives different results and even requires a measurement of the traffic density for the road
in scope. An easier way to define the required lighting class for a road is to classify roads
into different ‘zones’ (ex.: city center, shopping area, residential, industrial, parking and
so on). Each ‘zone’ may be assigned with a corresponding lighting class. To find these
assignments, Eandis supports the cities/communities with a ‘master plan public lighting’.
9
This master plan results in a clear view on a city’s requirements for the public lighting.
Each a road in scope is then classified to a zone and each zone is linked to the required
lighting level (lighting class).
DSOs do not track the full road configuration
Most asset managers of a public lighting installation (e.g., Eandis) track data of luminaries,
type of lamps, power, mounting heights and others but in general they do not record data
related to the geometry of luminaries arrangement like road width, position of a luminary
wrt the road, fixture inclination, lamp spacing and similar. To collect these data, Eandis
started using Mobile Mapping technology. The mobile mapping gives us the missing inputs
for the required lighting calculations.
High performance lighting calculation software is the key for optimization
When performing lighting calculations, software is needed to calculate the lighting result
from the different inputs. Most lighting calculation software on the market give DSOs the
possibility to check a result of a single configuration per an executed calculation task.
Some software have a sort of solution wizard, making calculations between fixed limits
(e.g., PhoCa software mentioned above), but actually we are still searching, as the DSO,
for the lighting calculation software that works with an input of the mobile mapping (i.e.,
an extended lighting infrastructure database). Today DSO Eandis needs to execute a mass
of separate calculation tasks using standard lighting calculation software to find the
optimizations of public lighting installations. There is a huge demand for lighting
calculation software which quickly compare different configurations in a flexible manner
(some limits fixed, e.g., spacing, mounting height), others not (e.g., luminary, type of
lamp, power).
8. Conclusions And Future Work
The high performance software (in terms of low computation times) for photometric
computations is the highly demanded tool supporting the redesign of public lighting
towards power consumption reduction. The prototype software, PhoCa, was created in
AGH University of Science and Technology. The test showed that it allows for obtaing 5-8%
reduction of the power usage.
Further development of the software focuses on implementing automation and
visualization in the process described in Section 3. As a proof of concept a prototype tool
has been implemented for 3D visualization. It is an extension to Maya rendering and
animation software. It mainly improves transition 3 (see Figure 3) by integrating
photometric calculations with the rendering engine. It is showed in action in Figure 5. The
scene consists of a flat urban area with four lamp posts. At each lamp post there is a
luminary (a light point) with given parameters. While the rendering engine shows how the
scene would look like photo-realistically, the photometric engine indicates underexposed
and overexposed regions (underexposure at the outer rim).
Furthermore, the proposed extension will be capable of calculating and optimizing
luminary parameters, minimizing or maximizing given criteria function e.g. power
consumption, public safety, overexposure etc. It can also optimize number of light points
or their distribution, proposing corrections to the design. The proposed extension is highly
interactive. While changing light point parameters, the over and under-exposure is
10
interactively calculated and visualized in real time. There will be no need to switch back
and forth between photometric calculation tool and 3D visualization one anymore.
References
[1] A. Sędziwy, M. Kozień-Woźniak, Computational Support For Optimizing Street Lighting
Design, Complex Systems and Dependability, vol.170, Advances in Intelligent and Soft
Computing, Ed. W. Zamojski et al., Springer Berlin Heidelberg, 2012
[2] Boyce, P., Hunter, C., Vasconez, S., An Evaluation of Three Types of Gas Station
Canopy Lighting. Technical Report. Lighting Research Center, Rensselaer Polytechnic
Institute, Troy, NY 12180-3352, 2001
[3] IESNA Lighting Handbook, Illuminating Engineering Society of North America (Author),
Mark Stanley Rea (Editor), 9th Edition, 2000
[4] Duco Schreuder, Outdoor Lighting: Physics, Vision and Perception, Springer Science +
Business Media B.V., 2008
11