Autonomous Street Lighting System based on Solar

Energy and LEDs M. A. Dalla Costa, L. Schuch, L. Michels, C. Rech, J. R. Pinheiro and G. H. Costa*

Federal University of Santa Maria – UFSM

GEPOC - GEDRE

Av. Roraima Nº 1000, CEP 97105-900

Santa Maria - RS – Brasil

* Universidade de Caxias do Sul – UCS

Centro de Ciências Exatas e Tecnologia – CCET

Rua Francisco Getúlio Vargas, 1130 - CEP 95070-560

Caxias do Sul - RS – Brasil

marcodc@ieee.org

Abstract – This work presents an autonomous street lighting

system based on solar energy as primary source, batteries as

secondary source, and lighting emitting diodes (LEDs) as light-

ing source. This system is being presented as an alternative for

remote localities, like roads and crossroads. Besides, it presents

high efficiency, because all power stages are implemented in DC

current. The design of LEDs fixture, in order to replace a 70W

high pressure sodium (HPS) lamp, is performed. This design

takes into account the human eye response in scotopic condi-

tions. LEDs driver and battery charger experimental results are

presented. The battery charger presents three control modes:

maximum power point tracker (MPPT) mode; constant current

mode; and constant voltage mode. The control mode depends on

the battery state (charged/discharged), and solar irradiance

level. The battery charger input impedance is analyzed in order

to ensure that the MPPT is obtained for any solar irradiance and

panel temperature.

I. INTRODUCTION

Nowadays, the power plants in Brazil are predominantly

based on hydroeletric energy. This happens because of the

high hydroelectric potential of the country. However, due to

the country continental dimension, the energy distribution

system is very huge and expensive. And, besides this huge

energy distribution system, there are a lot of localities without

energy supply [1]-[2].

The objective of this work is to develop an autonomous

street lighting system for remote places, with high efficiency

and long lifetime, using alternative energy. Besides, this solu-

tion can also be adopted as an alternative for the conventional

street lighting systems in urban centers.

Solar-electric-energy has grown consistently by 20%-25%

per year over the past 20 years, which is mainly due to the

decreasing costs and prices. This decline has been driven by

1) efficiency increase of solar cells; 2) manufacturing tech-

nology improvements; and 3) economies of scale [3]. In the

case of an autonomous street lighting system, the best solu-

tion is the solar energy option, because of the long lifetime,

easy installation, and modularity.

The main lamp types used in street lighting are the high

pressure discharge lamps, e.g.: mercury vapor lamps, HPS

lamps, and metal halide lamps. The discharge lamps demand

a ballast that provides their starting and steady state behavior,

which is commonly electromagnetic [4].

Lighting emitting diodes (LEDs) are being presented as an

alternative to replace the conventional lighting systems. Be-

sides their use as signalizing systems is much broadcasted,

they are not commonly used as lighting systems. However,

recent technology is improving gradually the LEDs efficiency

and color quality, which allows their application in lighting

systems [5].

The main advantages of using LEDs in the proposed light-

ing system are: their long lifetime (100,000 hours) that is

compatible with the solar panels lifetime (higher than 25

years); and their DC supply, exempting the use of an inverter,

which improves the circuit efficiency and decreases its cost.

Some examples of LEDs applied in street lighting can already

be found in the literature [6].

Figure 1 presents the proposed system that is composed by

a photovoltaic solar panel in order to charge the batteries

during the day through the DC/DC converter. This converter

is controlled by an MPPT algorithm. During the night, the

batteries supply the LEDs lamp through the driver, providing

the adequate current value and dimming the lamp when ne-

cessary. It can be observed that the whole system works in

DC, avoiding the energy wasting with additional inverter

power stages.

II. PROPOSED LIGHTING SYSTEM - LEDS

The number of LEDs used in the fixture is an extremely

important variable to be considered in the project of the light-

ing system, once both the designs of the photovoltaic solar

panels and battery bank are dependent of this number. The

number of LEDs needed in the fixture depends on the follow-

ing factors:

• Lamp to be replaced;

• Relationship between photopic and scotopic hu-

man vision;

• LEDs model.

Figure 1. Proposed system

978-1-4244-5697-0/10/$25.00 ©2010 IEEE 1143

A. Lamp to be Replaced

Public illumination systems are frequently based on high

pressure sodium (HPS) lamps instead of LEDs. For this rea-

son, a clear design methodology for LEDs based systems

cannot be found in the literature. In this work, a design me-

thodology is proposed, where the equivalence between a

standard illumination system and the LEDs based one is per-

formed. As a case study, a 70W HPS lamp was replaced,

specifically an OSRAM lamp model Vialox Nav-E Standard,

by a set of LEDs that leads to the same luminous efficiency.

An approximated spectral power distribution of the lamp is

illustrated in Figure 2. Other important characteristics of the

used model are:

• Luminous flux = 5600lm.

• CRI ≤ 25.

• Average life = 28000 h.

• Fixture efficiency = 80%.

B. Photopic Vision versus Scotopic Vision

The human eye has two types of photoreceptors in the reti-

na, rods and cones, responsible for sending visual signals to

the brain. In high levels of light, daylight, for example, the

cones are the majority photoreceptors, qualifying this vision

as photopic. At low light levels, the rods are the majority

photoreceptors, qualifying this vision as scotopic. In an in-

termediate light level, people deal with the called mesopic

vision, where both cones and rods are responsible for light

perception. Figure 3 presents the human eye sensitivity as a

function of the light wavelength, for the photopic and scotop-

ic conditions [7].

The current system of photometry, which determines the

luminous flux of light sources, is based on the photopic vi-

sion. In other words, the lamp to be replaced, characterized in

the previous section, has 5.600 lm (nominal flux) in photopic

vision conditions. However, when people deal with public

illumination of remote localities, as the application discussed

in this work, the scotopic vision characteristics are more ade-

quate to be considered. The perceptual equivalence of a no-

minal (photopic) luminous flux in a specific visual condition

is called effective flux or effective lumens [8].

Figure 2 –Spectral power distribution of the OSRAM Vialox Nav-E Stan-

dard.

Figure 3 – Sensitivity of the human eye as a function of the light wavelength,

under photopic and scotopic conditions.

C. LED Model Selection

Nowadays, a large variety of high power LEDs is commer-

cially available and, therefore, a careful study is required

before a specific model could be chosen. The most common

method to obtain white light is by using a blue LED coated

with phosphor that, when excited by the blue light, emits a

broad range spectrum, producing the white light. By this

method, it can be obtained LEDs of colors known as cool

white, neutral white and warm white, by varying the amount

of phosphor. The cool white LEDs are considered the most

efficient in scotopic conditions, since they require fewer

phosphor and produce light with wavelengths close to the

peak of sensitivity of the human scotopic vision. By this way,

cool white LEDs were chosen.

It is also important to notice that LEDs are commercially

available for 350, 700 and 1000mA (nominal mean current).

From a brief market analysis, it was noticed that the 700mA

LEDs certainly yield the lowest price per lumens, among the

high power LEDs. By this way, the model Luxeon Rebel –

Cool White Lambertian – 145 lm @ 700 mA was chosen. An

approximation of its spectral power distribution is illustrated

in Figure 4. Other important characteristics of this model are:

• Luminous flux = 145 lm.

• CRI ≥ 70.

• Average life = 50000 h.

• Fixture efficiency = 100%.

• Average power = 2.4 W.

Figure 4 – Spectral power distribution of the Luxeon Cool White Lambertian

LED.

1144

D. LEDs Arrangement Design

The efficiency of a light source can be evaluated by the

power spectrum distribution of this source (figures 2 and 4)

weighted by the human visual spectral sensitivity function

(Figure 3). As it was discussed in Section II.B, the human

visual perception depends on the lighting and viewing condi-

tions (photopic or scotopic), and the nominal luminous flux of

a source (the standard commercial characteristics) is usually

determined considering the photopic human visual sensitivity.

In this section, the number of LEDs is calculated based on the

efficiency of the lamp to be replaced and on the efficiency of

the previously modeled LEDs, both in scotopic conditions.

The efficiency of a light source under specific lighting

conditions can be evaluated by integrating the power spec-

trum distribution weighted by the human visual spectral sensi-

tivity function [9]. This methodology leads to a figure of

merit that is called source effective illumination:

��������, ��� ��� � � ����� ��, �������, ��� ( 1)

where cond is the lighting condition (photopic or scotopic),

source identifies the light source, λ is the wavelength,

P(source,λ) is the power spectrum distribution function and

V(cond,λ) is the human visual spectral sensitivity function for

a specific lighting condition.

Notice that the power spectrum distribution characteristics,

when informed by the manufacturer, usually correspond to a

family of sources, which can assume distinct nominal power

and luminous flux. Considering it, these characteristics are

usually available in a normalized form, as illustrated in fig-

ures 2 and 4. Thus, it is necessary the power spectrum distri-

bution to be weighted by a constant that leads to the nominal

luminous flux value (effective illumination under photopic

conditions).

The effective illumination relation, of the lamp to be re-

placed, between scotopic and photopic conditions, calculated

using (1), leads to a value approximately equal to 0.73. This

value means that a source model Vialox Nav-E Standard has

73% of its nominal efficiency when used in scotopic condi-

tions, which represents a source effective illumination of

approximately 4088 lm. This same relation, when the LEDs

modeled in Section II.C are considered, leads to a value of

approximately 2.23, which represents an efficiency in scotop-

ic conditions 2.23 times greater than in photopic (nominal)

conditions. This result shows that this model of LEDs reaches

approximately 323 effective lumens in scotopic conditions.

To replace the Vialox lamp, observing the lighting condi-

tions of the application considered in this work, the effective

lumens in scotopic condition should be 4088 lumens. Thus,

13 units of the chosen LEDs are enough to reach this lumin-

ous flux. Considering that LEDs have 20% of its initial (no-

minal) light decreased during its average life, the number of

LEDs must be raised to 15 units. Finally, since nominal val-

ues of a LED are obtained under a junction temperature of

25ºC and they use to decrease when temperature increases,

resulting in 20 units of LEDs (30% of safety margin) in the

proposed fixture.

III. LEDS DRIVER DESIGN

The LEDs driver design depends on the input voltage,

which is supplied by the batteries, being 24 V; the number of

LEDs (output power) and their connection too. In section II it

was defined that 20 LEDs are enough for the proposed sys-

tem. LEDs series connection was chosen due to its simplicity

and low cost. Then, the design parameters are listed below:

• Input: 24 Vdc.

• Output: 68 Vdc, 700 mA, 47.6 W.

• Switching frequency: 50kHz.

• Input current ripple: 25%.

The selection of the power converter that will perform the

LEDs supply is based on the design parameters above.

Among the possible power converters available, the most

indicated is the boost converter, due to its simplicity and low

cost. This power converter can be observed in Figure 5.

Figure 5 – LEDs driver – boost converter (50W@50kHz)

A. Experimental Results

A laboratory prototype was implemented, in order to prove

the theoretical analysis presented in former sections. The

design equations for this converter are well known in the

literature and the driver parameters, according to Figure 5, are

presented in Table I.

Table I – LEDs Driver Component Values

Vbat 24V - 2 x 12MF63

LD 640µH, 80 turns on core EE30/14

CD 1µF / 100V polypropylene

MD IRF530 – VDS = 100V, ID = 14A

DD MUR120

LEDs 20 x Luxeon Rebel – Cool White – 145 lm @ 700 mA

Figure 6 shows the converter input voltage and current,

which corresponds to the batteries output voltage and current

waveforms. Figure 7 shows the LEDs lamp voltage and cur-

rent, where it can be observed that the LEDs instantaneous

current is below 1 A, which is the maximum value recom-

mended by the manufacturer.

Figure 6 – Batteries voltage (UP) and current (DOWN)

(10V/div; 1A/div; 10µs/div).

1145

Figure 7 – LEDs lamp voltage (UP) and current (DOWN)

(20V/div; 0,5A/div; 10µs/div).

In order to perform the power measurements, the current

sensing resistors were short-circuited. Then, Pin = 47.6 W and

PLEDS = 44.2 W and the converter efficiency was 93%.

IV. BATTERY BANK

Taking into account the costs of each component, the bat-

tery bank becomes the most onerous in a photovoltaic system

[14], representing as much as 15% [15] of the initial costs for

installation of the photovoltaic system, or even up to 46%

[15] if maintenance costs are taken into account. This ex-

pense increase is mainly due to the lower batteries lifetime,

compared to the other system components. In cyclic applica-

tions, where the battery is charged and discharged daily, the

battery is the most expensive element of the system through-

out its useful life.

The recommended charge method is one current and one

voltage level (IU) as shown in Figure 8. The current value has

direct influence in the charge time, which should be up to

20% of the capacity of the battery (CAh) being charged.

In this specific application, and taking into account the

changes of solar radiation during the day and, consequently,

the energy available, the battery bank charge algorithm will

present three different modes:

• If the panel available current is lower than 0.1CAh, the

converter will find the maximum power point (MPPT)

in order to provide the highest current possible for the

battery bank (area under the current level, see figure 8).

• If the panel available current is higher than 0.1CAh, the

converter limits the current to 0.1CAh, and disables the

search for the maximum power point.

• If the batteries are already charged, the control algorithm

will apply a constant voltage level, in order to keep the

batteries charged.

Figure 8. Proposed charge method.

The MPPT chosen algorithm was Disturbance and Obser-

vation [16]-[17]. Figure 9 shows the implemented battery

charge control algorithm. This algorithm is composed of three

control loops:

- Current Loop: PI regulator with current reference of

0.1CAh (5A);

- Voltage Loop: PI regulator with voltage reference of

2.2V/cell (26.4V);

- MPPT Loop.

Besides, inside the voltage loop, the battery current is also

evaluated, in order to detect any problem during this loop

(e.g.: damaged battery). If the current is higher than 0.1CAh, a

protection subroutine is activated.

Figure 9. Battery charger control algorithm.

V. PHOTOVOLTAIC PANEL AND BATTERY BANK DESIGN

The main parameters used in the design of the panel and

the battery bank are shown in Table II.

Table II. Main design parameters

Autonomy taut = 12h

Maximum power P = 50W

Battery charger efficiency ηbat = 90%

Led driver efficiency ηD = 90%

Photovoltaic panel correction factor CF = 0.9

Solar radiation R = 3.57kWh/m2

Battery bank voltage Vbat = 24V

Initially, daily consumption should be defined, which for

this application was defined as 450W, where a dimming of

50% of the power in half the time was considered. The pro-

posed dimming is due to the reduction in man’s activities

during the period from 11 PM to 5 AM. Using the methodol-

ogy presented in [17], it has been defined that a 205Wp panel

is adequate for this application.

Table III. KD205GX-LP specifications of the Photovoltaic panel

Power 205Wp

Voltage at maximum power (Vmp) 26.6V

Current at maximum power (Imp) 7.71A

Start

Basic configuration

Acquisition of Vbat

and Ibat

Vbat>V

float

yes

Voltage loop

Vsetpont

= 2.2V/C

Ibat>0.1C

20

yes

Ibat>0.1C

20

yes

Current loopIsetpont

= 0.1C20

MPPT

Duty Cycle and

PWM atualization

1

1

1146

The capacity in Ah of the battery bank is defined by the

equation (3), and results in a minimum capacity of 23.62Ah.

However, as the discharge capacity should not exceed 30%,

the minimum capacity moves to 59.05Ah. By this way, two

63Ah (12MF63) Moura Clean batteries connected in series

were chosen.

��� � ����� � 56724 � 23.62Ah ( 2)

VI. BATTERY CHARGER DESIGN

The converter design for charging the battery bank and

finding the maximum power point (MPPT), basically depends

on the voltage provided by the photovoltaic panels, which is

around 26.6V to the maximum power point (Vmp). It also

depends on the current and the voltage of the battery bank

charge, which will have a maximum charge current of 5A

(0.1CAh) and bank nominal voltage is 24V for this application.

The design parameters are listed as follows:

• Input: 26.6 Vdc.

• Output: 24 Vdc, 5A, 120 W.

• Converter frequency; fSW = 20 kHz.

• Output voltage ripple: 0.5%.

As the output voltage is similar to the input voltage, the

battery charger topology should be a voltage step up/down

converter. Besides, with the increase in the panel’s tempera-

ture the output voltage is reduced. In this way, in order to

guarantee the correct working function of the system with a

variation of temperature, the non-isolated Cuk converter was

chosen (see figure 11) due to the following characteristics:

• Output voltage may be lower or higher than the

input (buck-boost characteristic);

• The input and output current are continuous.

PV ++

Lc1

Mc

Cc1

Dc

Lc2

Cc2

VBat

Figure 11 – Battery charger - Cuk converter (200W@20kHz).

The design equations of this converter can be checked in

[10] and the converter component values are presented in

Table IV.

A. Cuk Converter Input Impedance Analysis

The maximum power point operation is obtained when

there is impedance matching between the photovoltaic (PV)

module and the Cuk converter. As the output impedance of

the photovoltaic module has variations for different solar

irradiations and temperature levels, the input impedance of

the converter have to be adjusted. The input impedance of the

Cuk converter can be modified by the duty-cycle (d) as fol-

lows:

%&'�, �� �()*)+

2,-./012 31 5 1�6 , 8 1

1 5 1�

����1 9 �22

11 9 1 9 � , : 1

1 5 1� ;

(3)

where fsw is the switching frequency, rbat is the battery resis-

tance and m = Vbat/Vpv (battery voltage / PV panel voltage).

Figure 10 shows the input impedance of the Cuk converter

for extreme operating conditions of PV and battery bank

(Tables II and III). As can be seen, the input impedance can

be adjusted from 0 to ∞ ohms in all operating points. There-

fore, the proposed charging method shown in Fig. 8 ensures

the maximum power point tracking for all load conditions.

Figure 10 was obtained considering the battery and PV

nominal voltages (m=0.902), minimum battery voltage and

maximum PV panel voltage (mmin=0.767), and maximum

battery voltage and minimum PV panel voltage (mmax=1.2).

Figure 10. Cuk converter input impedance in CCM and DCM modes.

B. Experimental Results

A close loop laboratory prototype was implemented, in or-

der to prove the theoretical analysis presented in former sec-

tions.

Table IV – Battery Charger Component Values

VPV 26.6V – KD205GX-LP Photovoltaic panel

LC1 1.4 mH, 75 turns on core EE55/21

LC2 1.4 mH, 75 turns on core EE55/21

CC1 150µF / 250V electrolytic

CC2 33µF / 250V electrolytic

MC SPW47N60C3 Cool MOS Power Transistor

DC MUR830 – VRRM=300V, IF=8A

Vbat 24V - 2 x 12MF63

Figure 12 presents the Mosfet MC voltage (DOWN) and

current (UP). Figure 13 presents the PV panel voltage (UP)

and current (DOWN). Figure 14 presents the batteries voltage

(UP) and current (DOWN). All waveforms were acquired

during the current loop (Vpv=26.6V and Vbat=28V), where the

small low frequency ripple in voltage and current waveforms

can be denoted, like the adequate charge current value.

Figure 12 – Drain-to-source voltage (DOWN) and current (UP) across MC

(50V/div; 5A/div; 10µs/div).

1147

Figure 13 – PV panel voltage (UP) and battery current (DOWN)

(10V/div; 2A/div; 100µs/div).

Figure 14 – Battery voltage (UP) and battery current (DOWN)

(20V/div; 2A/div; 100µs/div).

In order to perform the power measurements, the current

sensing resistors were short-circuited. Then, Pin = 139.5 W

and Pout = 120 W, resulting in a converter efficiency of 86%,

during the current loop mode. The converter efficiency under

voltage loop mode is always higher, due to the small conduc-

tion losses in this mode.

VII. CONCLUSION

This work has proposed an autonomous street lighting sys-

tem, which uses solar energy as primary source, batteries as

secondary source, and LEDs as lighting source. This system

is an interesting solution for remote localities, as for roads

and crossroads.

The system presents high efficiency, since all the power

stages are DC-DC. This kind of conversion yields an easy

implementation and control.

The LEDs technology has been significantly improved in

the last few years, and they have been considered a promising

alternative to the illumination systems. The main advantages

of using LEDs are: high average life; high luminous efficien-

cy; and simple drives, when control and dimming systems are

required.

The photopic and scotopic human sensitivity were consi-

dered on the design of the lighting system. It was shown that

a 70W high pressure sodium lamp (model OSRAM – Vialox

Nav-E Standard) can be substituted by 20 LEDs with 145 lm

@ 700 mA (model Luxeon Rebel – Cool White Lambertian),

resulting in the same effective lumens.

The batteries are the components with highest cost in the

system. The main reason is its low average life: usually less

than four years. Depending on the operation conditions, the

average life of the batteries can still be strongly reduced.

The battery bank charge method is with one level of vol-

tage and current. The maximum current never can exceed

0.1CAh. The disturbance and observation method, to find the

maximum power point of the solar panel, has been imple-

mented when the available energy is lower than 0.1CAh. The

battery charger input impedance was analyzed in order to

ensure that the MPPT is obtained for any solar irradiance and

panel temperature.

Finally, experimental results of LEDs driver and battery

charger were presented, in order to validate the proposed

system.

ACKNOWLEDGMENT

The authors would like to acknowledge the University of

Caxias do Sul and Intral S/A for funding.

REFERENCES

[1] "Inserção das Energias Alternativas Renováveis no Setor Elétrico

Brasileiro," 3º Fórum Brasileiro de Energia Elétrica, 2003. (in

Portuguese) [2] "Programa Luz para Todos," in Ministério de Minas e Energia

(www.mme.gov.br), 2005. (in Portuguese)

[3] J.M Carrasco, et. Al., “Power-Electronic Systems for the Grid Integration of Renewable Energy Sources: A Survey.” IEEE Trans-

actions on Industrial Electronics, vol. 53, no 4, pp. 1002-1016, June

2006. [4] R.E. da Costa, J.S. da Silveira, F.L. Tomm, T.B. Marchesan, A.

Campos, R.N do Prado, “Warm-Up and Steady-State Control of

High-Pressure Sodium Lamps Applied to Public Lighting Systems.” Industry Applications Society Annual Meeting (IAS), 2008.

[5] A.Zukauskas, M.S. Shur, R. Coaska. Introduction to Solid-State

Lighting, Ed. Wiley Inter-Science, 2002. [6] White Paper: Street Lighting. Philips, Lumileds.

[7] Y. He, M. Rea, A, Bierman, J. Bullough, “Evaluating Light Source

Efficacy Under Mesopic Conditions Using Reaction Times”, in Proc of Illuminating Eng. Soc. Of North America Annual Conf.,

pp.236-257, 1996.

[8] Ian Lewin, "Lumen Effectiveness Multipliers for Outdoor Lighting Design", Journal of the Illuminating Engineering Society 30, pp.

40–52, 2001.

[9] T. Goodman, A. Forbes, H. Walkey, “A Practical Model for Mesop-ic Photometry”, in Proc of XVIII IMECO, pp.1-6, 2006.

[10] N. Mohan, T. M. Undeland, W. P. Robbins, Power Electronics: converters, applications, and design, John Wiley & Sons, 2nd Edi-

tion, New York, USA, 1995.

[11] A. Luque, G. Sala, G.L. Araujo, T. Bruton, “Cost Reduction Poten-tial of Photovoltaic Concentration”. International Journal of Solar

Energy, v. 17, p. 179 - 198, 2006.

[12] Available: www.cresesb.cepel.br/potencial_solar.htm [13] IEEE recommended practice for sizing lead-acid batteries for

photovoltaic (PV) systems 16 March 2001.

[14] S. Duryea, S. Islam, W. Lawrance, “A battery management system for stand-alone photovoltaic energy systems”, IEEE Ind. App.

Magazine, Vol. 7, Issue: 3, pp. 67-72, 1999

[15] IEA, “Management of Storage Batteries used in Stand-Alone Pho-tovoltaic Power Systems – Report_IEA_PVPS_T3-10:2002”, Inter-

national Energy Agency (IEA) 2002

[16] K. C. Oliveira, M. C. Cavalcanti, G. M. S. Azevedo, and F. A. S. Neves, "Comparative Study of Maximum Power Point Tracking

Techniques for Photovoltaic Systems," in VII INDUSCON, 2006.

[17] V. Salas, M.J. Manzanas, A. Lazaro, A. Barrado, E. Olias, "The control strategies for photovoltaic regulators applied to stand-alone

systems" ECON 02 Annual Conference of the Industrial Electronics

Society, IEEE, Volume 4, pp. 3274-3279, 2002 [18] Liu, Y.-H.; Leou, R.-C.; Cheng, J.-S.; "Design and Implementation

of a Maximum Power Point Tracking Battery Charging System for

Photovoltaic Applications" IEEE Russia Power Tech, pp. 1-5, 2005.

1148