Individual power supply to nodes in a wireless sensor network in a greenhouse using photovoltaic modules

TVE-F 18 006

Examensarbete 15 hp Juni 2018

Individual power supply to nodes in a wireless sensor network in a greenhouse using photovoltaic modules

Johannes Dufva

Timmy Mattson Lindgren

Teknisk- naturvetenskaplig fakultet UTH-enheten

Besöksadress:

Ångströmlaboratoriet Lägerhyddsvägen 1 Hus 4, Plan 0

Postadress:

Box 536 751 21 Uppsala

Telefon:

018 – 471 30 03

Telefax:

018 – 471 30 00

Hemsida:

http://www.teknat.uu.se/student

Abstract

Individual power supply to nodes in a wireless sensor network in a greenhouse using photovoltaic modules

Johannes Dufva, Timmy Mattson Lindgren

This thesis investigated the possibility of integrating a small photovoltaic module in a wireless sensor network node prototype made for use in crop production, mainly in greenhouses. The main question was if the simple photovoltaic module could provide enough power to the prototype's recharge system in order to continuously recharge the battery and thereby reducing the time maintaining the device due to its power consumption. Through measurements, both of the energy supplied by the potential photo voltaic modules and the prototype's power demand, the conclusion was that the power would not be sufficient due to the concealing environment in which the device would be placed. However, suggestions for further work was given in how the proposed idea could be developed.

Ämnesgranskare: Martin Sjödin Handledare: Nils Weber

Populärvetenskaplig sammanfattning

Tre olika typer av solceller har undersökts i växthusförhållanden. Målet har varit att avgöra möjligheten att ladda batterier med solpaneler som i sin direkta närhet är anslutna till en mätnod. Mätnoden är en del av det system som det nystartade företaget Bitroot AB utvecklar, vilket ska förenkla och automatisera bönders arbete i växthus. När detta skrevs användes ett 2000mAh, 3.7V litiumjonbatteri för mätnodens energiförsörjning. För att ladda batteriet användes laddningselektronik med en brytström på 10% av batteriets kapacitet. På grund av brytströmmen bör strömmen som ska levereras till laddningskretsen ligga på 20% av batteriets kapacitet.

Arbetet utgick från att solpanelen konfigurerades på sitt enklaste sätt. Detta betyder att celler seriekop- plas till dess en önskad spänning uppnåts och att celler sedan parallelkopplas till dess en önskad ström uppnåts. Mätningar visade att 15.66% av den solstrålning som träffar växthuset strålar ner till solpan- elens tänkta position. Detta resulterade i att den genererade strömdensiteten i solcellerna blev för liten och att den totala solpanelen var tvungen att vara så stor att det skulle hämma användarvänligheten och följaktningsvis också bli dyr. Det cellmaterial som trots detta presterade bäst i experimenten var kristalint kisel.

Det finns dock fler saker att undersöka innan arbetet läggs ner. Först bör batteriet bytas ut mot ett med lägre kapacitet. Detta skulle innebära att strömmen som levereras till laddningselektroniken inte behöver vara över 400mA. Därefter bör laddningselektroniken bytas ut, alternativt förenklas kraftigt, så att brytströmmen kan sänkas. På så sätt kan strömmen som levereras till laddningselektroniken sänkas ytterliggare. Vidare rekommendationer till fortsatt arbete var att undersöka möjligheten att parallelkoppla celler för att leverera en hög ström, som sedan kan omvandlas till spänning. Viss ström måste dock ledas direkt förbi omvandlaren för att leverera laddningselektroniken med den ström som fortfarande krävs. En fjärde rekommendation var att bygga ett stativ för modulen för att höja den över växterna. Stativet måste dock vara lätt att montera för att inte hämma användarvänligheten.

Acknowledgement

We would like to express our gratitude to our supervisor Nils Weber at Bitroot AB for his valuable input and generous help throughout the project. We would also like to thank Uwe Zimmermann at the Department of Solid State Electronics at Uppsala University for helping us with the measuring equipment and our speculations. Lastly, we would like to thank our project mentor Martin Sjödin for helping us with the finishing touch on the report and for his comments that inspired us along the way.

Contents

1 Introduction 5

1.1 Background . . . 5

1.2 Objective . . . 5

2 Theory 5 2.1 Photovoltaic effect . . . 5

2.2 I-V characteristics . . . 6

2.3 The solar spectrum . . . 8

2.4 Quantum efficiency . . . 9

2.5 Photovoltaic cell materials . . . 9

2.6 Power supply to WSNs. . . 9

2.7 Bitroot AB’s WSN . . . 10

2.8 Charging WSN node’s battery. . . 11

3 Method 11 3.1 Quantum efficiency . . . 11

3.2 Spectrum measurement . . . 12

3.3 I-V characteristics . . . 13

3.4 Node prototype measurement . . . 14

4 Results 14 4.1 Quantum efficiency and spectrum . . . 14

4.2 I-V characteristics . . . 16

4.3 Node prototype . . . 17

4.4 PV module sizing. . . 19

5 Discussion 19 5.1 Power demand observations vs. measurements. . . 19

5.2 Measurements. . . 20

5.3 PV module sizing. . . 20

5.4 Suggested further work. . . 21

6 Conclusions 21

7 References 21

Glossary

WSN Wireless sensor network.

PV Photovoltaic.

IR Infrared.

Bandgap EgBandgap is an unique attribute for a PV cell which refers to the energy difference between the valence band and the conduction band.

MP Maximum power.

MPP Maximum power point.

FF Fill factor.

STC Standard test conditions.

AM Air mass.

IQE Internal quantum efficiency.

EQE External quantum efficiency.

SR Spectral response.

RF module Radio frequency module.

RX/TX mode Receive/transmit mode for the radio frequency module.

aSi Amorphous silicon.

CIGS Copper indium gallium selenide.

cSi Crystalline silicon.

CCCV Constant Current/Constant Voltage.

0.1C A charge rate. Equivalent with the current required to fully charge the battery in 10 (= 0.1⁻¹) hours.

1 Introduction

1.1 Background

Nodes in Wireless Sensor Networks (WSN) today often run on batteries that frequently need to be recharged or changed completely. In order to minimize the human effort put into maintaining a func- tioning WSN in a cost effective and environmentally friendly way, this thesis aims to investigate whether it is possible to put small photovoltaic (PV) modules in direct contact with each node, thus continuously charge their battery. The nodes in question allow for a change in the configuration, i.e. the sensor setup for a given node can be changed at any time. A change in configuration may cause a change in power demand of that node. The nodes are easy to move if another measurement location is desired, thus the PV module’s size must not be too large.

If the node is supposed to be placed directly onto the soil, the circumambient plants will be covering the PV module and thereby reducing the amount of photons reaching it from the sun. Due to the plants green color, mostly wavelengths in the green part of the visible spectrum and wavelengths in the infrared (IR) spectrum will be transmitted. Modules with a small bandgap, required to absorb photons in the IR spectrum, might turn out to supply a voltage too small for the node’s demand. On the other hand, modules with a high bandgap, required to absorb photons in the green spectrum with a high energy, might turn out generating a too small current density. The PV module’s performance under these conditions needs to be compared with the node’s power demand and the charging specifications in order to find a good fit.

The idea behind this project began with the company Bitroot AB, which is constructing a WSN in order to measure moisture, temperature, nutrients and light conditions for individual plants in a greenhouse.

If the conditions for each plant in a greenhouse are known, the need for natural resources (e.g. water or plant nutriment) may be reduced drastically.

1.2 Objective

The project’s objective was to produce a preliminary study for a potential energy supply model for Bitroot’s WSN node prototype. Moreover, the goal of the preliminary study was to answer whether the method of using PV modules to supply the nodes with power in a greenhouse works without interfering with the applicability and further, find an optimal PV cell material for the particular application. This means that this thesis is a theoretical model for Bitroot’s self-charging WSN system. This thesis could further be used as a recommendation in a future project in which the goal is to produce a power supply prototype.

2 Theory

2.1 Photovoltaic effect

According to modern physics and quantum physics in particular, light consists of photons with energy depending on its frequency, ν, which is described by the relation,

E = hν,

where h is Planck’s constant. In some cases, it is convenient to express the photon energy in terms of the wavelength λ. Using the relation ν = _λ^c, where c is the speed of light, the photon energy E can be expressed as,

E = hc

λ, (1)

explicitly showing that the photon energy is inversely proportionally to the wavelength. One of the benefits of the understanding of the property of light is PV cells, that convert the solar energy to electricity in semiconductors. This conversion is possible due to the photoelectric effect. The bandgap E_g is a property of the semiconductor material and an energy separation between the valence band E_v and conduction band E_c of the PV cell such that E_g = E_c − E_v. In order for an electric current to be generated, electrons in the valence band must be excited to the conduction band, creating electron hole pairs in the conduction band. To avoid recombination, i.e. no electrons pass through the load, the cell is doped with n-type and p-type material which forces the holes and electrons in different directions [1]. In accordance with the photoelectric effect, the energy of the photon must be greater or equal to the bandgap energy. However, if E > Eg still only one electron hole pair is generated and the excess energy will generate heat. These two factors of energy loss, i.e. E < Eg produce no electron hole pairs and E ≥ Eg produce one electron hole pair, are the single most important limitations to a PV cell’s efficiency and is called the ultimate efficiency. In 1961, Shockley and Queisser calculated the ultimate efficiency for a single p-n junction PV cell at 0K that absorbs energy from the sun approximated as a black body to be approximately 44% for silicon with bandgap Eg= 1.1eV [2]. From this, it follows that for a relatively high bandgap less photons will get absorbed, i.e. the current density decreases, but with high energy and for a relatively low bandgap more photons will get absorbed, producing a higher current density, but at a lower energy. From this reasoning, it is clear that the output voltage is proportional to the band gap energy, and the relation is given by,

V ≈Eg

2q, (2)

where q is the elementary charge [3].

2.2 I-V characteristics

The maximum theoretical power i.e. the nominal power that a PV cell can produce is given by the short circuit current multiplied with the open circuit voltage (PN OM = ISCVOC). However, as the current at VOC must be zero and the voltage at ISC must be zero in actual conditions, this gives that the power at these points will always be zero. In other words, in actual conditions when the PV cell is connected to a load, the I-V curve can not be rectangular. This can be seen in Fig. 1 where both the nominal power PN OM and the actual maximum power PM P are indicated.

P

P

V

V

I

I

I

V

Figure 1: Illustration of the nominal power point (VOC, ISC) and the maximum power point (VM P, IM P).

The different powers are equivalent to the area of the rectangles, and the fill factor (FF) is the ratio of the areas _P^{PM P}

N OM.

Instead, the point on the I-V curve that produces the maximum power (MP) is defined as the maximum power point M P P ≡ (I_{M P}, V_{M P}) and the ratio of the areas _P^{PM P}

N OM is called the cell fill factor (FF) [4].

This results in the actual power produced by the PV cell is given by the equation,

M P ≡ F F · PN OM = PM P. (3)

The related efficiency of the PV cell is given by,

η =PM P

S , (4)

where S is the total power of the incident light. The curve I(V ) behaves as the diode equation when illuminated. Assuming that superposition principle works, I(V ) in Fig. 1 is given by,

I = I₀(e^nkTqV − 1) − I_ph, (5)

where n is the diode quality factor typically between the values 1 and 2, k is the Boltzmann constant, T is the temperature in the junction and Iph is the photon generated current that drives the load. I0 is a current-constant determined by the PV cell and is given by,

I0= I00e^−nkTEg , (6)

where I₀₀is a constant [4]. Eq. 5implies that the short circuit current,

ISC = −Iph (7)

i.e. the short circuit current is directly proportional to the light intensity. For VOC ≈ 0.5, and _kT^q ≈ 40 at room temperature, Eq. 5 can be approximated to,

0 = I0e^qVOCnkT − Iph. (8)

From Eq. 8, VOC = ^nkT_q ln(^I_I^ph

0 ), showing that VOC depends logarithmically on the light intensity. The temperature dependence of VOC can be derived by substituting Eq. 6 in Eq. 8,

V_OC= E_g q +nkT

q ln(I_ph I00

). (9)

I00 >> Iph, so ln(^I_I^ph

00) < 0 [3]. VOC decreases with an increasing temperature. Eq. 5 implies no temperature dependence for ISC.

The same rules also apply for cells connected in series or parallel. However, in cells connected in series the same current passes but the voltage drop increases according to the superposition principle. Meanwhile, cells connected in parallel will produce a higher current according to Kirchhoff’s law but the voltage drop will remain the same. Eq. 7 and Eq. 9 implies that the open circuit voltage is not as sensitive as the short circuit current under decreased light intensity. The desire to assemble the energy supply system in direct contact with the node may cause problems such as partial shading of one cell in the module.

Partial shading of one cell in a string of cells in series will reduce the total current to the shaded cell’s generated current as the same current must pass through all cells.

2.3 The solar spectrum

When choosing a PV cell for a setup, it is important to also know the spectral irradiance. The spectrum is typically presented in three ways: power P per wavelength λ and square meter m²; number of photons per wavelength λ, square meter and second s · m²; or number of photons per eV, square meter and second s · m².

To commercially compare the performance of PV cells, standard test conditions (STC) are used. The standards used are 1000W/m², 25^◦C and air mass (AM) 1.5[5]. Air mass is the path the photon take relative Earth’s surface normal. AM1.5 occurs when the incident light is at an angle of approximately 48^◦ [6]. Fig. 2 illustrate AM0 and AM1.5 radiation, where AM0 radiation is the radiation measured outside the atmosphere.

300 400 500 600 700 800 900 1000 1100

Wavelength [nm]

0 50 100 150 200 250

Intensity [ W/cm

nm]

AM0 radiation AM1.5 radiation

Figure 2: Illustration of the spectral irradiance presented in power per wavelength and square meter.

The figure also show how the spectral irrandiance changes depending on the air mass. Data retrieved from NREL [7].

In this thesis STC are not used as the spectral irradiance was measured in a greenhouse and was different from Fig. 2. This further means that the efficiencies of the PV cells will be different from the efficiencies featured in the their corresponding datasheets. The total power per square meter S in Eq. 4is given by the integral over all wavelengths.

The Swedish institute of meteorology and hydrology (SMHI) has a database of the measured global solar radiation in Sweden between years 1983 and 2014. The part of the year when the WSN system will be in use is according to Bitroot between February and October. The mean power during this period scaled to mW/cm² is 12.12 mW/cm² [8].

2.4 Quantum efficiency

There are two types of quantum efficiencies, internal quantum efficiency (IQE) and external quantum efficiency (EQE). The IQE is the ratio between absorbed photons and the incoming photons. The EQE is the ratio between the successfully collected charge carriers and the incoming photons. For this reason, this thesis concerns only the EQE [9]. The EQE depends on the wavelength and is given by the equation,

EQE(λ) = E(λ)

q · SR, (10)

where E(λ) is given by Eq. 1and SR is the spectral response of the PV cell. In order for the EQE to be calculated, the SR must first be measured. The result is a measure of how well the PV cell collects charge carriers for specific wavelengths in the solar spectrum. If multiplied with the spectral irradiance scaled to photon flux, the short circuit current density Jsc can be calculated as the integral over all wavelengths.

The equivalent is true if the spectrum in intensity is multiplied with the EQE and the resulting product is scaled to photon flux and then integrated over all wavelengths.

2.5 Photovoltaic cell materials

There are many different semiconductor materials used in PV cells. This thesis will investigate a few of the commercially available materials such as amorphous silicon (aSi), copper indium gallium selenide (CIGS) and crystalline silicon (cSi). PV cells are commonly divided into two categories; cSi and thin film PV cells [10]. The market is dominated by cSi PV cells, presumably because of their high efficiency and stability [4][10]. The most important among thin film PV cells are aSi, Cadmium telluride (CdTe) and CIGS. Thin film PV cells are characterized by better performance under bad light conditions and in higher temperatures, but their STC efficiency are worse than cSi PV cells.

2.6 Power supply to WSNs

Among the components making up a WSN node, usually three are defined as the main parts: the sensor or sensors used in measuring different types of data (e.g. temperature, humidity, light etc.), the microcontroller i.e. the node’s computer containing the CPU, memory and other peripherals and lastly, the radio frequency (RF) module used for transceiving data between the base station, other nodes and itself [11].

When supplying a node with power from an integrated PV module, the module first needs to transfer its output power to a rechargeable unit which then powers the node. The main problems lie in choosing the PV module that provides sufficient output voltage and current, choosing the right type of battery with suitable voltage constrains (min and max voltage when recharging it) and transferring the voltage and current from the PV module to the battery in a regulated way.

Since the objective of this project was not to produce the actual power supply prototype but instead investigate if such a prototype could be made having sufficient efficiency in supplying power to Bitroot’s

node prototype, the focus will be on choosing the right PV cell. Still, this problem is indirectly impacted by how well the other problems are solved later on.

2.7 Bitroot AB’s WSN

The WSN node which this thesis concerned was a prototype made by Bitroot AB. The unique attribute in this node was the proposed function (not yet integrated) of allowance for a change in the configuration, i.e. the ability to change the sensor setup. Thereby, calculating the maximum power consumption of such a modular node must be done with a full setup with all available sensors.

The calculation of the power consumption was done using the given values (seen in Tab. 1 and 2) in each of the component’s supplied datasheet and was then simulated for a theoretical prototype with a MATLAB script. The values obtained were between typical and maximum values and based upon the component’s reference designs. The duty cycle in which the node first wakes up, does a "test run"

(listens in receive (RX) mode, measures and then sends data in transmit (TX) mode) and then goes to sleep again was set to a period of 20 seconds i.e. one test run each 20th second.

Table 1: Parameter specification for the sensors and the microcontroller used in Bitroot’s node prototype from supplied datasheets [12] [13] [14]. Values used were between the typical and maximum and for the reference designs.

Component

Parameter

VCC (V ) IDrawn(mA) PDrawn(mW ) tSampling(ms) Temperature and hu-

midity sensor

2.4 - 5.5 0.8 - 1.5 2.64 - 8.25 2.5 - 15

Light sensor 1.7 - 3.6 0.00065 - 0.0016 0.0011 - 0.00576 100 - 800

Microcontroller 1.8 - 3.6 6.9 - 7.3 18.5 - 26.28

(in run mode), 0.0023 - 0.0119 (in standby mode)

2.2 - 3.0 (both wake-up-from- standby and run time)

In Tab. 2, an example message was used to calculate the time used by the RF module to send in TX mode. The message was simulated as if the temperature/humidity sensor would send 5 bytes and the light sensor 7 bytes of data. Besides the data, the message also consisted of a 4 bytes preamble, 4 bytes sync, 9 bytes header, 8 bytes CPU status and 2 bytes Cyclic Redundancy Check (CRC). This made up a total of 39 bytes or 312 bits for the message. It was sent at a frequency of 433 M Hz and with a bitrate of 4.6 kbps.

Table 2: Parameter specification for the RF module used in Bitroot’s node prototype from supplied datasheet [15] at 433 M Hz and with a bitrate of 4.6 kbps. Values used were between the typical and maximum and for the reference designs.

Component Parameter

VCC (V ) ISleep

(µA)

IRXmode

(mA)

IT Xmode

(mA)

PRX

(mW )

PT X

(mW )

tEx.message

(ms) RF module 1.8 - 3.6 0.2 - 100 15.0 - 17.0 13.1 - 16.0 27.0 - 61.6 23.6 - 43.2 67.8

Comparing the values for the sensors and the microcontroller in Tab. 1 with those for the RF module in Tab. 2, the power needed for receiving and transmitting messages for the RF module is substantially higher than that for the sensor measuring and processing in the microcontroller. This is recurrent for most WSN nodes [11].

Some tests with Bitroot’s node prototype in a greenhouse had been made previous to this thesis. The duty cycle had at first been set to 15 minutes and the prototype had then been left in the greenhouse to measure for over ten days before shutting down due to power loss. Thereafter the duty cycle had been changed to 10 seconds whereon the prototype now had lasted for three days before shut down.

2.8 Charging WSN node’s battery

With a device that need a specific supply voltage, its connected battery has to have a marginaly higher voltage. Further, a PV module used to then charge the battery needs an even higher voltage, which has to be considered when sizing the PV module. Bitroot’s node prototype that this thesis focus on is supplied with a 3.7V nominal voltage 2000mAh lithium-ion battery cell. The typical max charge voltage of batteries with nominal voltages in the range 3.6V-3.7V is 4.2V [16]. Additional voltage may be needed as an input voltage to a charging unit which is recommended for safe recharging. The reference charging device used in this thesis is based on a TP4056 IC [17]. The required input voltage is 5V.

The TP4056 charges the battery with Constant Current/Constant Voltage (CCCV), which is a charge cycle composed of three main stages. Constant current, where the battery capacity is charged until battery voltage reach saturation. Constant voltage, where the battery is charged with constant voltage, 4.2V, until capacity reaches 100% or the current reaches 0.1C, i.e. the current required to fully charge the battery in 10 hours. After the CV stage, the charge terminates to avoid overcharging [18].

Depending on how fast it is desired to charge the battery, the CC stage can be chosen differently.

However, as the charging terminates when the current decreases below 0.1C it is important to design the current higher than 0.1C. Therefore a good range would be 0.2C-1C.

3 Method

In this thesis three different PV cells were investigated: cSi, aSi and CIGS. At first, in order to get a correct bandgap energy, measurements of the cells EQE was performed at Ångströmslaboratoriet, Uppsala University in the department of Solid State Electronics. Secondly, measurements of a greenhouse spectrum with less and more concealment from plants were made, first in a home made environment on the roof of Ångströmslaboratoriet and later also in an actual greenhouse in the town Sala. Further, the I-V characteristics were also measured for each PV cell in the earlier mentioned home built environment and in the greenhouse. Lastly, a measuring of the current consumption and supply voltage was done for a Bitroot node prototype to compare with the theoretical power consumption values previously done with MATLAB.

3.1 Quantum efficiency

The quantum efficiency was measured between 300 − 1100nm as this was the wavelength range of rel- evance. The setup consisted of six main components: a light source, a chopper working at the same frequency as the amplifier to filter out surrounding/unwanted light, a monochromator used to send out nearly monochromatic light over the whole desired wavelength range, a test board where the PV cells were placed and lastly a current amplifier to fortify the signal and a lock-in amplifier to compare the signal with the reference and obtain the actual quantum efficiency. The actual setup can be seen in Fig.

3.

Figure 3: Quantum efficiency measurement setup with 1. Light source, 2. Chopper, 3. Monochromator, 4. Reference output/Light transfer tunnel, 5. Test board and 6. Current amplifier/Lock-in amplifier

Before measuring the three PV cells, a calibration was made with two PV cell samples with known quantum efficiency spectrum.

3.2 Spectrum measurement

The spectrum was measured with a spectroradiometer LI-1800, seen in Fig. 4a.

Sensor

Spectroradiometer LI-1800

(a)

(b) Figure 4: Spectroradiometer LI-1800 and measuring in the greenhouse.

The LI-1800 is capable of measuring radiation from 300−1100nm. The step size used was 2nm. Radiation in three different light conditions were measured. The first spectrum was measured under short tomato plants in a greenhouse in Sala, see Fig. 4b. The sensor was placed near the soil under the green leaves in a position where the prototype node might be set. In order to compensate for sudden intensity changes due to cloud illumination, the spectrum was measured as a mean value of ten different measurements under a shorter period of five minutes. The second spectrum was measured in a home built environment

to mimic the radiation when the plants are more grown, see Fig. 5. Material used for the simulated greenhouse was chicken wire and spinach leaves. Lastly, a spectrum of direct sun light was also measured.

Figure 5: Measurement setup in the home built environment.

3.3 I-V characteristics

The I-V characteristics of each solar cell were measured simultaneously as the spectrum in order to get the efficiency right. A simple circuit with two multimeter was designed with a resistive load, see Fig. 1.

Figure 6: Circuit used to measure I-V characteristics.

As a resistive load, a potentiometer was used for alterations in the magnitude of the load. The size of potentiometers used were (in Ω) 1k, 5k, 10k, 100k (the 1k was able to reach lower values of a minimum 0.9Ω). The use of four different potentiometers were due to hardship in doing small alterations at low resistance values with the larger potentiometers, but a large resistance value was still required for the measurements of VOC.

3.4 Node prototype measurement

The power consumption for a Bitroot node prototype was measured with an oscilloscope and two probes.

The first probe measured the voltage over a 100Ω resistor in series with the prototype’s battery and the main circuit, places at the low-side of the circuit. The second probe then initially measured the voltage directly over a LED light on the prototype’s main circuit, but was later moved to the Chip Select Not (CSN) pin. The purpose of the first probe was to measure how much current was drawn from the battery by the prototype. Moreover, the purpose of the second probe was to be used as a trigger for the measurements, i.e. to get the oscilloscope to start measuring when there was a voltage change over the main circuit. The setup was chosen for being simple yet still accurate [19]. It can be seen in Fig. 7.

Second probe

First probe

Figure 7: Setup for measuring Bitroot’s node prototype with an oscilloscope (blue unit), prototype (grey unit), a battery connected to the prototype and two measuring probes.

The node prototype was programmed to do a "test run" every 20 second, i.e. a duty cycle of 20 seconds.

The test run included the RF module waking up from sleep to first listen in RX mode, measure data with the sensors, then switch to TX mode and send data to its base station. Measurements were done for different setups (e.g. both or no sensor, RX off and TX on, LEDs connected to the microprocessor on or off) to analyze the power consumption for each component. The result was then to be compared with values in the datasheets [12] [13] [14] [15].

4 Results

4.1 Quantum efficiency and spectrum

The measured quantum efficiency for each of the three PV cells can be seen in Fig. 8. They are plotted both over the photon energy eV and the corresponding wavelength λ. From Fig. 8 it can be seen that aSi has the highest bandgap of approximately 1.77eV , CIGS the lowest of approximately 1.03eV and cSi a bandgap of approximately 1.07eV .

300 400 500 600 700 800 900 1000 1100 1200 Wavelength [nm]

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9

Quantum Efficiency

aSi CIGS Si 1.1 1.2 1.3 1.4 1.5 2.0

2.5 3.0 3.5 4.0

Photon energy [eV]

E_g≈ 1.03eV E_g≈ 1.07eV

E_g≈ 1.77eV

Figure 8: Quantum efficiency for aSi, CIGS and cSi with respect to wavelength and photon energy.

The spectrum was further measured in three different environments. The spectrum measurements can be seen in Fig. 9. It additionally includes the quantum efficiencies for each PV cell multiplied with the spectrum. This product was first scaled to photon flux and then integrated to find a value for JSC, see Tab. 3

400 500 600 700 800 900 1000

Wavelength [nm]

0 10 20 30 40 50 60

Intensity [µW/(cm2*nm)]

Measured spectrum aSi CIGS cSi

(a) Spectrum in home built environ- ment. Total power is 12.73 mW/cm².

400 500 600 700 800 900 1000

Wavelength [nm]

0 10 20 30 40 50 60

Intensity [µW/(cm2*nm)]

Measured spectrum aSi CIGS cSi

(b) Greenhouse spectrum. Total power is 22.62 mW/cm².

400 500 600 700 800 900 1000

Wavelength [nm]

0 20 40 60 80 100 120 140 160 180

Intensity [µW/(cm2*nm)]

Measured spectrum aSi CIGS cSi

(c) Spectrum under full sun. Total power is 81.28 mW/cm².

Figure 9: Spectrum from three different environments. (a) Home built environment consisting of leafs to block some of the radiation. (b) In a greenhouse. (c) Under full sun.

Table 3: Short circuit current density values. The values are the integral over all wavelengths of the product of the EQE and the spectrum.

Short circuit current density JSC [mA/cm²]

PV module

Fig.9

(a) (b) (c)

aSi 0.1 0.6 2.1

CIGS 6.6 9.7 33.8

cSi 7.4 11.0 38.6

From Fig. 9aand9cit was calculated that 15.66% of the power transmit through the leaves used in the home built environment. This, in combination with the data from SMHI, leads to an average of 1.90 mW/cm² transmit through the leaves in the home built environment during a full season.

4.2 I-V characteristics

The I-V curve together with the corresponding power curve for each PV cell that was measured, both in the home built environment and in the greenhouse, can be seen in Fig. 10. Additionally, the curve for the PV cells in full sun can be seen in Fig. 11. The values for current and power are normalized per area and the values for the voltage is per cell.

0 0.2 0.4 0.6 0.8

0 0.5 1 1.5

Current density [mA/cm2]

0 0.05 0.1 0.15 In home built environment 0.2

0 0.2 0.4 0.6 0.8

Voltage [V]

0 0.5 1 1.5

0 0.05 0.1 0.15 0.2

Power per area [mW/cm2]

In greenhouse

(a) aSi

0 0.1 0.2 0.3 0.4 0.5

0 2 4 6

Current density [mA/cm2]

0 0.5 1 1.5 2 2.5

Power density [mW/cm2]

In home built environment

0 0.1 0.2 0.3 0.4 0.5 0.6

Voltage [V]

0 2 4 6

0 0.5 1 1.5 2 In greenhouse 2.5

(b) CIGS

0 0.1 0.2 0.3 0.4 0.5 0.6

0 2 4 6 8

Current density [mA/cm2]

0 1 2 3

Power density [mW/cm2]

In home built environment

0 0.1 0.2 0.3 0.4 0.5 0.6

Voltage [V]

0 2 4 6 8

0 1 2 In greenhouse 3

(c) cSi Figure 10: I-V curves and the corresponding power curve for each PV cell.

0 0.2 0.4 0.6 0.8

0 0.2 0.4 0.6 0.8 1 1.2

Current density [mA/cm2]

0 0.05 0.1 0.15 0.2 0.25

Power density [mW/cm2]

In full sun

(a) aSi

0 0.1 0.2 0.3 0.4 0.5 0.6

0 5 10 15 20 25 30

Current density [mA/cm2]

0 2 4 6 8 10 12

Power density [mW/cm2]

In full sun

(b) CIGS

0 0.1 0.2 0.3 0.4 0.5 0.6

0 5 10 15 20 25 30 35

Current density [mA/cm2]

0 2 4 6 8 10 12 14

Power density [mW/cm2]

In full sun

(c) cSi Figure 11: I-V curves and the corresponding power curve for each PV cell.

Furthermore by interpreting Fig. 10 and 11, specific I-V characteristic values and measured efficiency values for each PV cell can be seen in Tab. 4 and5. All values are for a single PV cell and the current values are normalized per square centimeter cm². The measuring error was ±1.2% for current values and ±0.5% for voltage values (in temperature of 23^oC± 5^oC).

Table 4: I-V characteristic values and measured efficiency for each PV cell.

In home built environment

PV module

Parameter JSC

(mA/cm²)

VOC(V ) JM P

(mA/cm²)

VM P (V ) PM P

(mW/cm²)

Efficiency (%)

aSi 0.61 0.74 0.35 0.34 0.12 0.93

CIGS 5.07 0.50 3.96 0.40 1.57 12.34

cSi 6.14 0.52 5.56 0.40 2.22 17.46

In greenhouse

aSi 1.03 0.75 0.52 0.38 0.19 0.86

CIGS 5.67 0.52 5.36 0.42 2.23 9.85

cSi 7.37 0.54 6.87 0.37 2.53 11.18

Table 5: I-V characteristic values and measured efficiency for each PV cell.

In direct sun light

PV module

Parameter J_SC (mA/cm²)

V_OC(V ) J_{M P} (mA/cm²)

V_{M P} (V ) P_{M P} (mW/cm²)

Efficiency (%)

aSi 1.17 0.75 0.57 0.35 0.20 0.25

CIGS 27.84 0.53 24.58 0.41 10.03 12.34

cSi 32.16 0.55 26.64 0.48 12.71 15.63

4.3 Node prototype

The power consumption values of a test run with Bitroot’s node prototype from the oscilloscope can be seen Fig. 12, 13and14. In Fig. 12, two measurements can be seen for the node with the LED turned on and off, in Fig. 13two measurements with the sensors turned on and off (zoomed in on the first part of test run on channel one) and in Fig. 14one measurement with the RF module’s RX mode turned off.

Here channel one (yellow graph) corresponds to values from the first probe over the 100Ω resistor and channel two (purple graph) to values from the second probe over the main circuit (values were not used in results as the second probe only worked as a trigger).

(a) LED turned ON before sleep. (b) LED turned OFF before sleep.

Figure 12: Node prototype with LED turned ON and OFF before sleep.

(a) Sensors ON. (b) Sensors OFF.

Figure 13: Node prototype with sensors ON and OFF.

Figure 14: Node prototype with RF module’s RX mode turned OFF.

In Fig. 12the second probe was set to measure over the LED. Then, in Fig. 13and14, it was moved to the CSN which explains the disappearance of the purple graph. The voltage value over the main circuit measured with the second probe in Fig. 12was 1.59V in RX and TX mode an peaked to 3.71V when the LED was turned on. Further, the values for different the segments in the first probes measurements from Fig. 12can be seen in Tab. 6.

Table 6: Measured power consumption for Bitroot’s node prototype with duty cycle of 20 s.

First probe (yellow graph)

Segment

Parameter

Vmeasured(mV ) Imeasured(mA) tsegment(ms) RatioCycle(%)

Sleep 12.4 0.12 18.88 94.42

Idle 20 0.20 0.22 1.09

RX mode 158 1.6 0.82 4.12

TX mode 158 1.6 0.07 0.37

With the duty cycle of 20s used during the measurements, the average current becomes I_AV = 0.19mA.

Comparatively, I_AV dependent on different lengths of the duty cycle can be seen in Tab.7. The discharge rate for the lithium-ion battery used with the prototype is also stated.

Table 7: The presumed (with the datasheets) and measured average current and discharge rate dependent on the duty cycle.

Duty Cycle (s) IAV (mA) Rate of discharge (C)

Data sheet Measured Data sheet Measured

900 0.14 0.12 7e−5 6.2e−5

120 0.24 0.13 1.2e−4 6.7e−5

20 0.80 0.19 4e−4 9.4e−5

10 1.48 0.25 7.4e−4 1.3e−4

2 6.92 0.78 3.5e−3 3.9e−4

4.4 PV module sizing

As stated in Section2.8the required input to the charger unit must be 5V . The lowest measured value of VM P for each cell was used to calculate the required number of cells in series in order to deliver 5V . Tab. 4and5 read the lowest value for aSi to 0.34V, CIGS to 0.40V and cSi to 0.37V. This means that aSi need 15 cells, CIGS need 13 cells and cSi need 14 cells in series in order to deliver 5V .

The current that needs to be delivered to the charging unit during CC stage depends on the capacity of the battery. The Bitroot node prototype is assembled with a battery of 2000mAh capacity which results in a CC of 400mA. The further results in Tab. 8 depicts total areal demand per cell or per parallel connection of cells based on IM P values from Tab. 4 and5.

Table 8: Total areal demand per cell for each PV cell.

Areal demand per cell (cm²)

PV cell

Environment

Home built environment Greenhouse Full sun

aSi 1143 770 702

CIGS 101 75 17

cSi 72 59 15

5 Discussion

5.1 Power demand observations vs. measurements

By comparing the datasheet values in Tab. 1 and 2 with the measured values, the presumption that the RF module would have the highest power consumption was confirmed both by the results in Fig.

13, where the influence of the sensors was shown to be minute in comparison and in Fig. 14, where the RX mode turned off resulted in a great change in amplitude. The calculated time to send the example message in Tab. 2was also about the same as the one in Fig. 12(68.7ms calculated, 70ms measured).

The significant deviation between the power demand of the prototype first estimated with the datasheet values in Tab. 1and2and the actual power consumption measured in Tab. 6 with the oscilloscope was the current drawn in RX and TX mode. As seen in Tab. 7, the average current I_AV together with the discharge rate increases faster and grows significantly larger for shorter duty cycles with the datasheet values than those from the prototype measurement. An explanation for this could be that the RF module is used differently than that of the reference designs which the datasheet values are based upon. Also, as stated in Section2.7, the values from the datasheets were between the typical and maximum values, henceforth it does not confute the measured values lower than these.

In spite of the stated difference, the discharge rate discussed was a lot lower than 0.2C when the lithium- ion battery was used with the prototype. This means that during charge, we do not have to be concerned about the discharge rate being larger than the charge rate. However, it should be kept in mind that the battery is discharged constantly regardless of the weather. The results from Tab. 7 reads that,

theoretically, the risk of the battery discharging fully is not likely. Even with values from the datasheet, and a duty cycle of 20s, the expected lifetime for the battery is estimated to 104 days. Furthermore, the mean value of solar radiation from Section2.3 together the measured power during full sun in Fig. 9c, it is reasonable to expect full sun 14.9% of the time during February to October. Of the season’s 272 days, this means about 40 days of sun like in Fig. 9c. Then again, this is a very rough analysis but still gives an estimation as to how the weather and the power supply system will cooperate.

5.2 Measurements

Overall, the measurements of the I-V curve were not easy to perform and there exist many methods to measure I-V characteristics more efficiently than with a resistive load used in this thesis. The main advantage of the method is the simplicity, but in contrast the main problem was to measure the values close to ISC for cSi as it required a small resistance. This problem can be seen clearly in Fig. 11c. It can also be seen that the data does not quite reach the MPP, and therefore the measured value for cSi efficiency should be somewhat higher than what is presented in Tab. 5. The measurements also had to be performed on a day without cloud interruption as a shifting amount of incident photons drastically changed the amount of generated current, causing irregularities in the data. This problem is evident in Fig. 10c, greenhouse curve. Furthermore the systematic errors of 0.5% in voltage values and 1.2% in current values are in the context considered negligible in comparison to the random errors consisting of human errors. The values on the multimeters, the amperemeter in particular, did not stay on a fix value but instead varied slightly. Thus it was quite hard to pick an accurate value.

The spectroradiometer was borrowed from the department of Solid State Electronics in Uppsala Uni- versity and was said to be calibrated. Assuming this, in combination with the fact that spectrum data is less prone to be affected by human errors, the values for J_sc presented in Tab. 3 should be more accurate than those presented in Tab. 4and5. It is notable that all J_scvalues in Tab. 3are higher than those measured and presented in Tab. 4 and5, except for aSi in the home built environment and in the greenhouse. This analysis lead to a hint that there could be some room for higher Jscvalues for cSi and CIGS in Tab. 4and5, which will cause an increase in IM P and thus also in PM P and the efficiency.

As has already been said, all measurements were performed during good weather. Therefore, the results should not be considered typical during the season. They are however a good indication of how large the fraction of the direct radiation are in the position where the PV module might be placed. Also notable is that the measurements in the greenhouse were performed at the beginning of the season, when the plants were still young and small in comparison to what they will be during most of the season, see Fig.

4b. With this in mind, Fig. 9a, was considered the most reliable spectrum measurement.

5.3 PV module sizing

The sizing of each cell in the PV module is presented in Tab. 8. It is clear that assembling an aSi module to the recharge system is very inefficient in size terms. Regarding the areal demand of CIGS and cSi in full sun, they differ with only 2cm², or 11.7% in cell area. However, as the number of needed cells is lower for CIGS the total area paid differ with 11cm², or 5%. However, in the home built environment the difference was 23.2% and in the greenhouse environment 15.3%. This means that in all cases, it would pay off assembling a cSi module.

The shadowing of the cells were a major problem. From Tab. 8, the required increase in area to deliver the same amount of current was a factor of 1.63 for aSi, 5.94 for CIGS and 4.80 for cSi. The obvious consequence of the increase in area is the higher cost, but also a reduced applicability. The Bitroot node must be high in mobility, in order for the farmer to change the position easily and measure elsewhere.

The required area of the module could also be hard to fit under the plants which in turn could intervene with the farmer’s plant space. To overcome this problem, a solution would be to mount the PV modules on the roof of the greenhouse. In effect, the mobility advantage gained by wireless communication would be diminished by using wires for powering or charging the node.

5.4 Suggested further work

This thesis focused on the possibility to charge a battery in a greenhouse with a PV module in all its simplicity, i.e. use the battery and charging module already used by Bitroot and no electronics besides the recharger module. Undeniably, the results discussed in Section5.3are hard to defend from a business perspective. Too much power is lost for a great cost increase. With this said, during this thesis some ideas on potential solutions came to mind.

Firstly, the battery should be changed to one with a lower capacity. Using a battery with a capacity of 1000mAh would decrease the sizing of the PV modules by half, as they would not have to deliver 400mA to reach 0.2C but instead 200mA. On the other hand, the discharge rates would double causing a duty cycle of e.g. 10s to be perhaps too frequent according to datasheet values. In the case of measured values of the discharge rate, 10s would still work. In fact, the battery capacity could be decreased even more depending on how frequent it is desired to take measurements.

Secondly, one option would be to find a charging module with a termination current less than 0.1C. This would allow for a PV module design that aims for less than 0.2C without risking to terminate the charge despite disadvantageous weather. If combined with the first suggestion, the designer would have to keep in mind that not only does the discharge rate increase, but the charge rate decreases. The designer would therefore have to program a measuring interval not too frequent.

Thirdly, another option would be to use just one cell or equivalently a chain of parallel connected cells.

The PV module could be designed in a way such that it delivers a high current to be converted to 5V . However, current equal to the CC stage charge rate would have to bypass the converter as it is still needed during recharge. The major advantage of this configuration is that shadowing of one cell would not affect the entire module and thus power could be saved. This would lead to a smaller areal demand and a cheaper power supply design.

Lastly, one option would be to construct some sort of mechanical stand for the PV module to be mounted on, enabling it to be raised above shadowing vegetation and obstacles. The stand should be easy to install and uninstall, perhaps something similar to a telescope mount with the wires inside.

6 Conclusions

Based on the effects on applicability, a PV-based power supply system in its simplicity does not generate a sufficient amount of power to be worth assembling to the node. The errors that the measurements contain are in the context considered small and as they will not affect the full design a great deal. In other words, a potentially higher current due to measurement errors would most likely not decrease the areal demand by a considerable amount. However, as discussed in Section 5.4 there is work that can be done to improve applicability which possibly could make it more sufficient and should therefore be further investigated before the proposed power supply system is disregarded. Because of the time limitation these topics were excluded from this thesis.

In summary, it is too early to say whether the use of an optimized, PV-based power supply system to recharge the sensor nodes is sufficient or not. With the current components it is, in the authors’ opinion, a too expensive option which would lead to expensive nodes and in the long run cause Bitroot AB as a company to be less competitive.

7 References References

[1] Jäger Klaus. Isabella Olindo. H.M. Smets Anro. A.C.M.M. van Swaaij René. Zeman Miro. The working principle of a solar cell. In Solar Energy Fundamentals, Technology, and Systems, Delft

University of Technology 2014, pages 23–26. 2014.

[2] William Shockley and Hans J. Queisser. Detailed balance limit of efficiency of p-n junction solar cells. Journal of Applied Physics, 32(3):510–519, 1961.

[3] Uwe Zimmermann. Senior lecturer at department of engineering sciences, solid state electronics.

Lecture Slides.

[4] Bolko von Roedern. Photovoltaic materials, physics of. In Cutler J. Cleveland, editor, Encyclopedia of Energy, pages 47 – 59. Elsevier, New York, 2004.

[5] O. Bel Hadj Brahim Kechiche, M. Hamza, and H. Sammouda. Performance comparison of silicon pv module between standard test and real test conditions. In 2016 7th International Renewable Energy Congress (IREC), pages 1–6, March 2016.

[6] C. Riordan and R. Hulstron. What is an air mass 1.5 spectrum? [solar cell performance calculations].

In IEEE Conference on Photovoltaic Specialists, pages 1085–1088 vol.2, May 1990.

[7] ASTM International. Astm g173 - 03(2012). https://www.astm.org/Standards/G173.htm, 2012.

[8] SMHI. Solstrålning i sverige. https://www.smhi.se/kunskapsbanken/meteorologi/

solstralning-i-sverige-1.89984, Jun 2015. Revised: Jan 2018.

[9] W. Ananda. External quantum efficiency measurement of solar cell. In 2017 15th International Conference on Quality in Research (QiR) : International Symposium on Electrical and Computer Engineering, pages 450–456, July 2017.

[10] Lundberg Olle. Band gap profiling and high speed deposition of cu(in,ga)se2 for thin film solar cells, Oct 2003.

[11] M. Srbinovska, V. Dimcev, and C. Gavrovski. Energy consumption estimation of wireless sensor networks in greenhouse crop production. In IEEE EUROCON 2017 -17th International Conference on Smart Technologies, pages 870–875, July 2017.

[12] STMicroelectronics. Ultra-low-power 32-bit MCU Arm -based Cortex^R -M0+, up to 192KB Flash,^R 20KB SRAM, 6KB EEPROM, LCD, USB, ADC, DACs, Aug. 2015. Revised Sep. 2017.

[13] Maxim Integrated. MAX44009, Industry’s Lowest-Power Ambient Light Sensor with ADC. 160 Rio Robles, San Jose, CA 95134 USA, Apr. 2007. Revised Nov. 2013.

[14] Sensirion. Datasheet SHT3x-DIS. Laubisruetistr. 50 CH-8712 Staefa ZH Switzerland, Jan. 2011.

[15] Texas Instruments. Low-Power Sub-1 GHz RF Transceiver. Dallas, Texas 75265, Apr. 2007. Revised Nov. 2013.

[16] Battery University Group. Bu-303: Confusion with voltages. http://batteryuniversity.com/

learn/article/confusion_with_voltages, May 2017.

[17] NanJing Top Power ASIC Corp. 1A Standalone Linear Li-lon Battery Charger with Thermal Reg- ulation in SOP-8, 04 2008.

[18] Battery University Group. Bu-409: Charging lithium-ion. http://batteryuniversity.com/

learn/article/charging_lithium_ion_batteries, Apr 2018.

[19] D. C. Harrison, D. Burmester, W. K. G. Seah, and R. Rayudu. Busting myths of energy models for wireless sensor networks. Electronics Letters, 52(16):1412–1414, 2016.