electronics

Article
Bee Sound Detector: An Easy-to-Install, Low-Power, Low-Cost
Beehive Conditions Monitoring System
Dimitrios I. Kiromitis 1, *, Christos V. Bellos 1, * , Konstantinos A. Stefanou 1 , Georgios S. Stergios 1 ,
Thomas Katsantas 1 and Sotirios Kontogiannis 2

1 Lime Technology IKE, 45221 Ioannina, Greece

2 Laboratory Team of Distributed Microcomputer System, Department of Mathematics, University of Ioannina,
45110 Ioannina, Greece
* Correspondence: [email protected] (D.I.K.); [email protected] (C.V.B.);
Tel.: +30-26-5103-2295 (C.V.B.)

Abstract: One of the most significant agricultural tasks in beekeeping involves continually observing
the conditions inside and outside the beehive. This is mainly performed for the early detection of
some harmful events. There have been many studies on how to detect and prevent such occurrences
by performing periodic interventions or, when the frequency of such actions is hard to enforce, by
using sensory systems that record the temperature, humidity, and weight of the beehive. Nevertheless,
such methods are inaccurate, and their delivered outcomes usually diverge from the actual event
or false trigger and introduce more effort and damage. In this paper, the authors propose a new
low-cost, low-power system called Bee Sound Detector (BeeSD). BeeSD is a low-cost, embedded
solution for beehive quality control. It incorporates the sensors mentioned above as well as real-time
sound monitoring. With the combination of temperature, humidity, and sound sensors, the BeeSD
can spot Colony Collapse Disorder events due to famine and extreme weather events, queen loss,
and swarming. Furthermore, as a system, the BeeSD uses cloud logging and an appropriate mobile
Citation: Kiromitis, D.I.; Bellos, C.V.; phone application to push notifications of extreme measurements to the farmers. Based on achieved
Stefanou, K.A.; Stergios, G.S.;
performance indicators, the authors present their BeeSD IoT device and system operation, focusing
Katsantas, T.; Kontogiannis, S. Bee
on its advantages of low-cost, low-power, and easy-to-install characteristics.
Sound Detector: An Easy-to-Install,
Low-Power, Low-Cost Beehive
Keywords: beekeeping monitoring devices; IoT; low-power; embedded systems; distributed systems
Conditions Monitoring System.
and services
Electronics 2022, 11, 3152. https://
doi.org/10.3390/electronics11193152

Academic Editor: Dongkyun Kim

1. Introduction
Received: 9 September 2022
Accepted: 28 September 2022 The Internet of Things (IoT) industry is evolving at frenetic rates in the agricultural
Published: 30 September 2022 sector, offering new solutions and assisting in turning traditional problem-solving pro-
cesses into automated technologies. Internet of Everything, the next progression of the IoT,
Publisher’s Note: MDPI stays neutral
includes the concept of all-around things and transparent connectivity to the cloud. More-
with regard to jurisdictional claims in
over, current IoT trends favor processes or algorithms supported by cloud-provided data
published maps and institutional affil-
iations.
visualization and probing monitoring, as well as detection, prediction, and decision-making
with the use of data mining algorithms and machine learning models.
One of the most important sections of agriculture is the apiary industry, which is
threatened by many illnesses and harmful events [1]. Beehive harmful events can be
Copyright: © 2022 by the authors. categorized into six major categories, as shown in Table 1.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).

Electronics 2022, 11, 3152. https://doi.org/10.3390/electronics11193152 https://www.mdpi.com/journal/electronics

Electronics 2022, 11, 3152 2 of 13

Electronics 2022, 11, x FOR PEER REVIEW 2 of 14

Table 1. Beehive condition incident classification and current sensory technologies used by IoT
Table 1. Beehive
devices [2–6]. condition incident classification and current sensory technologies used by IoT de-
vices [2–6].
Category Description IoT Devices and Sensors Used
IoT Devices and
Category Description
Colony Collapse Disorder (CCD)
C1: CCD famine due to famine and lack of feeding Sensorsand
Cameras Used
weight scales [2]
Colony Collapse Disorder
resources (CCD)
(pollen)
C1: CCD famine due to famineExtreme environmental Cameras and weight scales [2]
and lack of feeding
resources (pollen)
conditions due to climate change. Temperature and
C2: Environmental
Extreme environmental conditions
Very low/high temperatures. humidity sensors.
factors
dueExtended
to climatelow/high
change. humidity
Very due Temperature andstations [3]
Meteorological
C2: Environmental to rainfalls/aridity
low/high temperatures. Extended humidity sensors. Meteorologi-
factors
low/high humidity due to rain- cal stations [3] scales
Weight
C3: Swarming CCD due to swarming
falls/aridity and microphones [4]
C4: Queen loss CCD due to queen mortality Weight scales-
C3: Swarming CCD due to swarming
andCameras
microphones [4]
experimentally, in
C5: Diseases/Parasites Varroa mite, bee Nosema fungus
C4: Queen loss CCD due to queen mortality -
some cases only [5]
Cameras experimentally,
Cameras, vibration in
sensors,
C5: Diseases/Parasites
C6: External attacksVarroa mite, bee Nosema
Mammals, wasps,fungus
hornets
some cases only [5]GPS [6]
gyroscopes,
Cameras, vibration sensors, gy-
C6: External attacks Mammals, wasps, hornets
roscopes, GPS [6]
The categories C3, C4, and C5 are considered to be the most important, as they include
The that
events categories C3, C4,
can cause theand
lossC5of are considered
an entire to be
colony, butthe most
also important,
because they as
arethey
the in-
ones that
clude events
can be that canMany
prevented. cause the loss of
studies havean entire
shown colony, but also
that these because
events they
can be are the ones
detected early using
that can be
sound prevented.
signals, images,Many studies[7].
or weight have shown that
Detecting andthese events can
preventing suchbeharmful
detectedevents
early could
using
make sound
apiarysignals, images,
practices moreor profitable
weight [7].while
Detecting and preventing
reducing the risk ofsuch
CCD. harmful events
could make apiarypresents
This paper practicesamorenewprofitable while BeeSD,
system, called reducingthattheallows
risk of CCD.
observation of the dom-
This paper presents a new system, called BeeSD,
inant conditions inside and outside the beehive while using that allows observation
innovative of low-power
the dom- and
inant conditions inside
low-cost architecture. and outside the beehive while using innovative low-power and
low-cost architecture.
The BeeSD system’s innovation includes continuous observation of beehives to detect
The BeeSD
harmful system's
events, and itsinnovation includes continuous
easy implementation makesobservation
BeeSD an of beehives to detect
easy-to-install system for
harmful events, and its easy implementation makes BeeSD an easy-to-install
every beekeeper. This goes beyond state of the art and includes a system coupled system for
with low-
every
cost, beekeeper.
low-powerThis goesand
sensors beyond state of the
peripherals, art can
which andidentify,
includesassess,
a systemandcoupled
cluster with
episodes of
low-cost, low-power
any of the sensorsinand
events shown peripherals,
Figure 1 [7,8]. which can identify, assess, and cluster epi-
sodes of any of the events shown in Figure 1 [7,8].

Figure
Figure 1. 1. BeeSD
BeeSD system
system high-level
high-level system
system architecture.
architecture.
Electronics 2022, 11, 3152 3 of 13

This paper’s structure is as follows: Section 2 presents related work, focusing on

existing IoT solutions for apiary incidents and their capabilities. Then, focusing on the
critical incidents mentioned in Table 1, the authors present their proposed solution of an
end-node device and system architecture in Section 3. Finally, Section 4 presents the authors’
system experimentation and validation, while Section 5 presents conclusions related to
their proposition.

2. Related Work
The use of IoT in beehive monitoring started about a century ago when Gates manu-
ally extracted temperature data and published his findings [9]. Since then, many studies
have been carried out to automate the conditions data logging procedures from beehives.
Moreover, with the evolution of technology in embedded systems, more and more parame-
ters have been added to the function of monitoring, with features such as humidity, gas
concentrations, and weight being some of them [8].
Since the early studies, research has shown that besides activity inside the beehive,
the environment greatly influences bee behavior. Thus, Wireless Sensor Networks (WSNs)
have been used to record environmental factors such as Temperature, Humidity, CO2 and
O2 levels, and dust, which may account for specific events (see Table 1, C2) [10,11]. Later
studies showed that weight monitoring can provide useful information about colony status
(Table 1, C1, and C3 cases) [12–14].
With the evolution of image processing, some studies have focused on detecting
diseases, parasites, or even external attacks (Table 1, C5, and C6 cases) using camera
modules [15,16]. However, despite research on these methods, sound processing seems to
dominate in the detection of many harmful events in apiaries, such as swarming, external
attacks, or even CCD (Table 1, C1, C3, C6 cases). As a result, researchers trying to detect
these events have focused on creating IoT devices that collect sound data [17–19].
In recent years, many systems have been created to combine the abovementioned
research, creating systems that collect different kinds of data simultaneously [20,21]. For
example, in [22], a new IoT beehive management system was presented to continuously
observe and monitor bee behavior to record colonies’ behavior during harmful events. In
addition, several systems have been developed to gather and process sensory data that
allow each node to wirelessly communicate with the main concentrator, which acts as an
Internet Gateway, sending the unprocessed data to remote database services [23].
Finally, several IoT market products help beekeepers to monitor their ‘beehives’ sta-
tus [24–26]. These products provide the potential to collect data and prevent harmful events
by installing sensor motes inside and outside the beehive while increasing the system’s
complexity. The proposed BeeSD system provides an easy-to-install device that is easily
installed in a beehive’s lid (or comes with a new lid with the device attached) without
harming or disturbing the bees. Section 3 describes the proposed system architecture
and components.

3. Proposed High-Level BeeSD System Architecture

The authors propose a new beehive conditions monitoring system called Bee Sound
Detection system, named for its core sensing functionality. The proposed BeeSD system
includes three stages, as shown in Figure 1, and includes the following components:
(1) the BeeSD end-node device, (2) the BeeSD concentrator, (3) the BeeSD cloud services, and
(4) the BeeSD monitoring mobile phone application.
The first stage of component functionality is the collection of temperature, humidity,
and sound data from the beehives through the implemented IoT BeeSD device (1). Next, the
end-node devices communicate using Wi-Fi with a central node device called a concentrator
(2). Each concentrator can support up to 10 devices simultaneously. Data are then sent to
and stored in the cloud (3) data-logging service via the concentrator to be cloud processed
(Figure 1(3)), which concludes with a decision about imminent harmful condition events
being sent via push notifications. BeeSD concentrator devices offer the ability to remotely
Electronics 2022, 11, 3152 4 of 13

monitor and control each one of the end-node devices over the cloud via an independent
VPN service channel per end-node device.
The authors utilized Wi-Fi high-bandwidth and low-power technology, among other
power-efficient technologies such as LoRaWAN and RF-based or direct NBIoT transmis-
sions. Moreover, the transfer of raw sound data cannot be easily performed over LoRaWAN
or RF-based technologies due to their narrow bandwidth, small data transmission payloads,
and frequency channel time-usage limitations [27]. In addition, using NBIoT technology
can achieve the required bandwidth for data uplinks. However, telecommunication costs
per beehive increase significantly according to existing telecommunication cost plans,
offering a monthly usage of 25 MB and 0.3–0.5 € per additional MB spent. For this rea-
son, a much cheaper concentrator device that utilizes a Wi-Fi to 3G/4G transponder for
cloud data transmissions was selected. If telecommunication costs drop in the following
years, the proposed architecture will be much easier to migrate transparently to Wi-Fi over
NBIoT concentrators.
The default measurement period (probing period) is 5 min per hour per end-node
device. However, the monitoring and probing service can adjust this period by utilizing
the power on/off actuator component. New on/off periods can be defined and uploaded
periodically to the end-node devices, with the on and off time intervals expressed in
minutes. A detailed description of the on/off interval setup is provided in Section 3.1.
The cloud server illustrated in Figure 1(3) controls the data processing and bee condi-
tions detection service, capable of covering and differentiating amongst condition cases,
as mentioned in Table 1. Finally, the beekeepers are alerted to harmful condition events
via the BeeSD mobile phone application through a cloud push notification service (4). The
beekeepers can also visualize real-time BeeSD measurements and trends (daily, weekly,
monthly) of temperature, humidity, sound intensity, and spectrogram frequency response,
i.e., most contributing frequencies provided by the FFT spectrogram analysis performed at
the BeeSD end-node device and uploaded to the cloud data-logging services.

3.1. BeeSD IoT Device Component

The authors propose a new end-node IoT device called the Bee Sound Detector device
to support their system’s distributed architecture. The proposed BeeSD consists of two
components, as shown in Figure 1: (A) the end-node device and (B) the photovoltaic (PV)
panel. The end-node device component includes: (1) main process unit, (2) microcontroller
unit (MCU) power controller, (3) humidity and temperature sensors, (4) lavalier micro-
phone, and (5) power actuator. The PV panel component includes (6) PV panel, (7) voltage
regulator, (8) pull-down converter (PDC), and (9) Lithium iron phosphate(LiFePO) battery.
All the abovementioned components are shown in Figure 2.
The BeeSD end-node IoT prototype includes the following:
1. The four-core ARM CPU (main processor unit), where the sensors are attached. The
card reader contains a MicroSD card of 32 GB, which contains the appropriate scripts
and storage for the extracted files.
2. A waterproof temperature sensor probe that returns the targeted temperature values
amongst frames inside the beehive.
3. The lavalier microphone that is used to record the sounds made by the bees, stored in
raw WAV data files.
4. The temperature and humidity sensor attached to the lid monitor the temperature
and the relative humidity values inside the beehive.
5. A micro-USB cable that used to power the ARM CPU via the PV panel component.
6. A power actuator that switches the device on and off as scheduled.
Electronics
Electronics2022, 11,11,
2022, x FOR
3152PEER REVIEW 5 of 14 5 of 13

Figure 2. BeeSD end-node IoT device components and parts.

Figure 2. BeeSD end-node IoT device components and parts.
To the end-node device component’s main processing board are attached the peripher-
alsTheforBeeSD end-nodeofIoT
the collection (3)prototype
temperature, includes the following:
humidity, and (4) sound data from the beehive, as
1. well
Theasfour-core ARM power
the (2) MCU CPU (main processor
controller thatunit), where
controls thethe sensors
power are attached.
on/off actuator. The
cardThe reader contains
authors’ a MicroSD
proposed BeeSDcardIoTof 32
mainGB, processor
which contains the appropriate
consists of an embedded scriptsfour-core
ARMand CPUstorage for 512
with the extracted
MB of RAM, files.wireless LAN, and Bluetooth transponders attached, and
2. a card
A waterproof
reader oftemperature
at least a 16sensor probe that
GB MicroSD returns
card. BeeSDthe targeted
requirestemperature
four cores and values
more than
512amongst frames inside the beehive.
MB RAM because sound processing necessitates these specs, and this can maintain
BeeSD as a low-cost system.
3. The lavalier microphone that is used to record the sounds made by the bees, stored
TheWAV
in raw PV panel
data component
files. includes a (6) PV panel 20 W/12 V that charges through a (7)
voltage regulator and the (9) battery. Finally, an (8) 12 V to 5 V PDC supplies the end-node
4. The temperature and humidity sensor attached to the lid monitor the temperature
device’s main processing ARM CPU unit.
and the relative humidity values inside the beehive.
Temperature and humidity sensors are connected to the ARM’s GPIO pins, whereas
5. A micro-USB cable that used to power the ARM CPU via the PV panel component.
the microphone is connected to it through the device’s USB port. The end node is placed
6. onAthe power actuator
beehive’s that
lid. switches
Every BeeSD thesystem
device on hasand off as scheduled.
a device working as the main concentrator,
to which all the other devices send their data. The board
To the end-node device component's main processing are attached
concentrator the is
device periph-
permanently
erals for the collection of (3) temperature, humidity, and (4)
connected to the cloud server through a VPN and continually uploads the data.sound data from the beehive,
as well as Thethepower
(2) MCU power (6)
actuator controller that controls
is connected to thethe
MCU power on/off
power actuator. and controls the
controller
The authors'
power feed ofproposed
the mainBeeSD IoT main
processor unit.processor
The MCU consists
power of an embedded
controller four-core
initially reads each
ARM CPU with 512 MB of RAM, wireless LAN, and Bluetooth transponders attached,
EEPROM start and stop interval value (in minutes) and controls the actuator accordingly
and a card reader of at least a 16 GB MicroSD card. BeeSD requires four cores and more
for a period of TP = Tstart + Tstop. It then waits for a command sent by the microprocessing
than 512 MB RAM because sound processing necessitates these specs, and this can main-
unit through its UART using the power serial protocol. If a command is received, then it
tain BeeSD as a low-cost system.
updates its start and stop intervals. This cycle is periodically repeated.
The PV panel component includes a (6) PV panel 20W/12V that charges through a (7)
Each of the main processor-attached sensors automatically records and sends its data
voltage regulator and the (9) battery. Finally, an (8) 12V to 5V PDC◦ supplies the end-node
device's sound
(raw data or FFT
main processing ARMresponse
CPU unit. textual data, temperature C, and humidity) to the cloud
orTemperature
information and management system after
humidity sensors the user has
are connected programmed
to the ARM's GPIO the scripts.
pins, The temper-
whereas
ature and humidity sensor is installed at the top of the beehive’s
the microphone is connected to it through the device's USB port. The end node is placed lid, while the waterproof
temperature sensor and the microphone are installed between
on the beehive's lid. Every BeeSD system has a device working as the main concentrator, the beehive’s frames.
to whichThe all end-node
the other devices
devicessend their data. using
communicate The concentrator
Wi-Fi with adevicecentralis communication
permanently and
connected to the cloud server through a VPN and continually uploads the data.
cloud upload node device called the BeeSD concentrator. Each concentrator can support up
to 10 end-node devices simultaneously. Data are then sent via the concentrator and stored
in the cloud (see Figure 1) data-logging service. Furthermore, BeeSD concentrator devices
offer the ability to remotely monitor each one of the end-node devices through the cloud
via an independent VPN service channel per end-node device. Furthermore, the cloud
Electronics 2022, 11, 3152 6 of 13

instantiates the data-processing and bee condition detection service capable of covering and
differentiating amongst condition cases, as mentioned in Table 1. Finally, the beekeepers
are alerted to harmful condition events via the BeeSD mobile phone application through a
cloud push notification service (see Figure 1(3)). The beekeepers can also visualize real-time
BeeSD measurements and trends (daily, weekly, monthly) of temperature, humidity, sound
intensity, and spectrogram frequency response. Most contributing frequencies are provided
by the fast Fourier transform (FFT) spectrogram analysis performed at the BeeSD end-node
device and uploaded to the cloud data-logging services.
Finally, the procedure becomes cheaper and faster as the processed data do not exceed
one percent of the unprocessed file size. Additionally, the new data reduce and minimize
energy consumption due to their low-energy components.
The BeeSD IoT device provides a revolutionary way of implementing its end-node
devices in the beehive, as it just replaces the simple lid that beekeepers tend to use. It is
also a low-cost device as it uses the newest ARM CPU, which provides four-core power to
process the data before sending it to the main concentrator. This method requires less data
consumption as both the temperature and humidity and sound data are transferred as CSV
or image files, with sound data converted to whatever is needed, using the power of the
four-core ARM CPU attached.

3.2. BeeSD Services and Service Capabilities

The BeeSD system, as described in Figure 1, includes the following services:
1. Audio data-logging service;
2. Sensory measurements logging service;
3. Synchronization service;
4. Monitoring and probing service;
5. Cloud processing service;
6. Push notification service;
7. Statistical trends service.
First, the audio data-logging service is responsible for sending the recorded raw data
to the cloud over FTP. In contrast, the sensory measurement-logging service (including FFT
results if textual sound preprocessing is selected or raw sound data) sends textual data
measurements over HTTP PUT. The stored WAV or CSV files are cloud processed, reaching
the result of the existence or not of the events listed in Table 1.
Along with the results from sound processing, there are some critical threshold values
of temperature and humidity to which beekeepers can be alerted by the push notification
service (Figure 3a) over HTTP JSON POST, providing information about unsuitable con-
ditions inside and outside of the beehives, as described in Table 1. Alert push messages
include temperature threshold values exceeding 35 ◦ C or below 10 ◦ C and relative humidity
values below 10% or above 90%.
NTP service synchronization over VPN is also implemented in BeeSD to keep the
devices synchronized with the time zone. This kind of synchronization provides the
capability to ensure that the recording session, depending on user preferences, works as
scheduled and is not affected by any kind of disruption, such as devices shutting down or
any sudden reboots.
The statistical trends service sends JSON temperature and humidity data over the
requested HTTP post time interval. Temperature data at the beehive lid (external temper-
ature) and inside frames (internal temperature probe) are illustrated in Figure 4 as mean
daily values for July 2022. Similar statistical trends for internal humidity values can be
shown for each BeeSD device by the BeeSD statistical trends Web panel.
 Electronics 2022, 11, x FOR PEER REVIEW 7 of 14

Electronics 2022, 11, 3152

conditions inside and outside of the beehives, as described in Table 1. Alert push messages7 of 13
include temperature threshold values exceeding 35 °C or below 10 °C and relative humid-
ity values below 10% or above 90%.

Figure 3. (a) Push notification service for unsuitable weather conditions, (b) Monitoring and probing
service User Interface.

NTP service synchronization over VPN is also implemented in BeeSD to keep the
devices synchronized with the time zone. This kind of synchronization provides the ca-
pability to ensure that the recording session, depending on user preferences, works as
scheduled and is not affected by any kind of disruption, such as devices shutting down or
any sudden reboots.
The statistical trends service sends JSON temperature and humidity data over the
requested HTTP post time interval. Temperature data at the beehive lid (external temper-
ature) and inside frames (internal temperature probe) are illustrated in Figure 4 as mean
Figure3.3.(a)
(a) Push
Push notification service for
forunsuitable weather conditions, (b) (b)
Monitoring and and
probing
daily values
Figure
service Userfor July 2022. Similar
notification
Interface. statistical
service trendsweather
unsuitable for internal humidity
conditions, values can
Monitoring beprobing
shown for each BeeSD device by the BeeSD statistical trends Web panel.
service User Interface.
NTP service synchronization over VPN is also implemented in BeeSD to keep the
devices synchronized with the time zone. This kind of synchronization provides the ca-
pability to ensure that the recording session, depending on user preferences, works as
scheduled and is not affected by any kind of disruption, such as devices shutting down or
any sudden reboots.
The statistical trends service sends JSON temperature and humidity data over the
requested HTTP post time interval. Temperature data at the beehive lid (external temper-
ature) and inside frames (internal temperature probe) are illustrated in Figure 4 as mean
daily values for July 2022. Similar statistical trends for internal humidity values can be
shown for each BeeSD device by the BeeSD statistical trends Web panel.

Figure 4. Temperature graphs of Jun 2022, taken by the statistical trends service User Interface
(Internal frame temperature, External Temperature in the beehive inner lid part).

Monitoring and probing services (see Figure 3b) allow users to know the status of
implemented devices and peripherals (online or offline). Besides this, this service can
set the on/off actuator intervals per beehive and provide real-time information for the
end-node sensors. Finally, this service allows users to choose the data they want to upload
(text or raw).
Regarding sound analysis and visualization, the input of the cloud-uploaded WAV
files from each beehive is split into minute raw sound data files. Next, for each of these
files, a low-pass filter is applied with a center frequency of fc = 2 kHz, focusing on the
sound frequency range produced by bees. A fast Fourier transformation (FFT) is then
applied per raw data minute file, producing an FFT spectrogram (FFT diagrams over time).
Figure 5a,b illustrates the original signals, with FFT accordingly, where the y axis contains
the amplitude measured in dBV (where dBV = −( 20 log mV V + 60 )).
 Regarding sound analysis and visualization, the input of the cloud-uploaded WAV
files from each beehive is split into minute raw sound data files. Next, for each of these
files, a low-pass filter is applied with a center frequency of fc = 2 KHz, focusing on the
sound frequency range produced by bees. A fast Fourier transformation (FFT) is then ap-
plied per raw data minute file, producing an FFT spectrogram (FFT diagrams over time).
Electronics 2022, 11, 3152 8 of 13
Figure 5a,b illustrates the original signals, with FFT accordingly, where the y axis contains
the amplitude measured in dBV (where 𝑑𝐵𝑉 = − ( 20 log + 60 ) ).

Figure 5. Representation of (a) signal during a minute, (b) frequency in the band of (0−2000 Hz), and
Figure 5. Representation of (a) signal during a minute, (b) frequency in the band of (0−2000 Hz), and
(c) aggregated frequency response (in 10 bins).
(c) aggregated frequency response (in 10 bins).

In addition, focusing on the Table 1 C3 and C4 cases, these values are narrowed down
In addition, focusing on the Table 1 C3 and C4 cases, these values are narrowed down
to the frequency
to the frequency range
range of of 300−600
300−600 Hz Hz andandthenthen classified
classified into into 10–25
10-25Hz Hz frequency
frequency bins bins
(300–325, 325–350, . . . , 57,5600). Each bin consists of a one-minute
(300–325, 325–350,..., 57,5600). Each bin consists of a one-minute sum of amplitudes for sum of amplitudes for
the specified frequency range (e.g., 200–235). The response to this
the specified frequency range (e.g., 200–235). The response to this classification process is classification process is
illustrated in Figure 5c. In swarming or queen loss events, there
illustrated in Figure 5c. In swarming or queen loss events, there is a significant response is a significant response
increase
increase(in(in
mV)mV) in in
thethemonitoring
monitoringfrequency
frequency binsbins
in that frequency
in that range.range.
frequency Appropriate
Appropriate
push
pushnotification
notification forfor
swarming
swarming or queen
or queen loss loss
is sent if at ifleast
is sent 5/10 bins
at least 5/10signify a 70% a 70%
bins signify
increase
increaseofoftheir
theiramplitude
amplitude values between
values between twotwo consecutive
consecutive probes.
probes.
Appropriate
Appropriate VPN
VPN service infrastructure
service infrastructure is set to monitor
is set to monitor BeeSD sensor
BeeSD and and
sensor nodenode
sta- status
tus
andand send
send push
push notifications
notifications wheneverany
whenever anyofof themdevelop
them developany anyminor
minororormajor
majorconnectivity
con-
nectivity or functionality
or functionality issues. issues. In addition,
In addition, through through
the VPN the VPN
probingprobing service,
service, beekeepers
beekeepers can check
can
andcheck andthe
change change
numberthe number
of probesof probes for asaslong
for as long the as the recording
recording duration.
duration.
The
Thebeekeepers
beekeepers cancanalso also
monitor all theall
monitor BeeSD devices devices
the BeeSD working working
under the under
same con-the same
centrator. However, as is explained in the next section, this affects
concentrator. However, as is explained in the next section, this affects the energy and datathe energy and data
consumption
consumption of of
thethe
system.
system. Thus, oneone
Thus, of the
of biggest
the biggest challenges
challengesof BeeSD is to maximize
of BeeSD is to maximize
the quality of data while keeping the energy footprint
the quality of data while keeping the energy footprint and data consumptionand data consumption as lowas aslow as
possible. The authors' experimentation and BeeSD system
possible. The authors’ experimentation and BeeSD system validation follow. validation follow.

4. BeeSD System Experimentation

The authors implemented an experimental BeeSD system setup to validate their
system, as shown in Figure 6a. The experimental setup included five end-node beehive
prototypes, one primary concentrator (Wi-Fi to 5G), and three battery-enabled solar panels
to turn the BeeSD end-node component into an autonomous one. Figure 6b illustrates the
internal parts of the end-node prototypes. Prototype sensory parts were mostly low-cost
IoT sensors, including the DS18B20 temperature probe, DHT11 humidity and temperature
sensor attached to the beehive lid, and an omnidirectional microphone with a frequency
range of 35 Hz–18 kHz and −30 dB sensitivity. The ARM CPU, Wi-Fi transponder, and
MCU power controller were included in the transparent IP-56 plastic case.
 totypes, one primary concentrator (Wi-Fi to 5G), and three battery-enabled solar panels
turn the BeeSD end-node component into an autonomous one. Figure 6b illustrates t
internal parts of the end-node prototypes. Prototype sensory parts were mostly low-co
IoT sensors, including the DS18B20 temperature probe, DHT11 humidity and temperatu
Electronics 2022, 11, 3152 sensor attached to the beehive lid, and an omnidirectional microphone with
9 of 13a frequen
range of 35 Hz–18 kHz and −30dB sensitivity. The ARM CPU, Wi-Fi transponder, an
MCU power controller were included in the transparent IP-56 plastic case.

(a) (b)
Figure
Figure6.6.(a)(a)
BeeSD
BeeSDsystem proof-of-concept
system setup
proof-of-concept and
setup end-node
and prototype.
end-node TheThe
prototype. right figure
right shows
figure the the
shows prototype ex
perimentation performed in the beekeeping station. (b) The BeeSD prototype sensory parts.
prototype experimentation performed in the beekeeping station. (b) The BeeSD prototype sensory parts.

The proof-of-concept
The proof-of-concept system devicesystem devicefor
was tested was
its tested for its operational
operational durability, da
durability, data
transmissions to the cloud (data delivery capabilities), and energy consumption. In addition, In add
transmissions to the cloud (data delivery capabilities), and energy consumption.
tion, the
the BeeSD end-node BeeSDand
sensors end-node
systemsensors and
services system
were servicesas
confirmed were confirmed
functional as functional
in field
operations, providing a Technology Readiness 7 (TRL-7) system prototype demonstration demo
field operations, providing a Technology Readiness 7 (TRL-7) system prototype
stration in an operational environment. As a result of these tests, three Key Performan
in an operational environment. As a result of these tests, three Key Performance Indicators
Indicators (KPIs), presented in Section 4.1, were defined to confirm that BeeSD is a low
(KPIs), presented in Section 4.1, were defined to confirm that BeeSD is a low-cost, easy-
cost, easy-to-install, and low-power system. The KPI expected values were set using re
to-install, and low-power system. The KPI expected values were set using references to
erences to the functional capabilities of existing autonomous system used by the beehi
the functional capabilities of existing autonomous system used by the beehive industry
industry (specifically GPS safety systems and weight scales).
(specifically GPS safety systems and weight scales).
4.1. BeeSD Key Performance Indicators
4.1. BeeSD Key Performance Indicators
BeeSD aims to become a low-power, easy-to-use, and low-cost system. To elabora
BeeSD aims to become a low-power, easy-to-use, and low-cost system. To elaborate
and prove that the BeeSD prototype implementation maintains those attributes, some K
and prove that the BeeSD prototype
Performance implementation
Indicators were determinedmaintains
from thethose attributes,
authors' some Keyexperime
proof-of-concept
Performance Indicators were determined from the authors’ proof-of-concept experimen-
tation and market research. The BeeSD system KPIs and their corresponding referen
tation and market research.
attributes The BeeSD
are presented in system
Table 2. KPIs and their corresponding reference
attributes are presented in Table 2.

Table 2. BeeSD Key Performance Indicators.

Key Performance Indicator In-Process KPI

KPI 1: Low energy Continuous operation of at least 90 days.
All components are placed on the lid, including the battery
(maximum of 10,000 mAh). Batteries with capacities higher
KPI 2: Easy to install than 10,000mAh cannot fit inside the beehive lid, and their
weight makes them hard to attach and operate (remove the
battery, charge it, and place it back to the lid case).
Scalability of 5–10 beehives per concentrator. Integration of
more than 10 beehives per concentrator (20–50) quadruples
KPI 3: Low cost the concentrator cost and doubles the cost of the required
battery and photovoltaic (PV) panel needed to maintain the
concentrator’s autonomous operation.
Electronics 2022, 11, 3152 10 of 13

The main concentrator used by the BeeSD system operates autonomously, with power
supplied by a 60 Ah/12 V battery connected to a 20 Wh/12 V solar panel. The concentrator
requires more energy than the other end-node devices because it works continuously,
receiving data from 5–10 devices and uploading the sensory data to the cloud.

4.2. BeeSD End-Node Device Power Consumption

Several experiments were performed to verify BeeSD end nodes as low-energy devices
to determine the least possible power needed. The first experiment was to disable every
power-consuming part of the ARM peripherals and GPIO port and calculate its consump-
tion. The results showed that the ARM cores while being idle with no active peripherals
(Wi-Fi, HDMI, LEDs, and USB ports), consumed around 100 mA, whereas upon activating
Wi-Fi and USB, this rose to 120 mA.
BeeSD, as presented above, uses USB connectivity for the microphone on a 5 V power
supply and GPIO connectivity for its sensors, which work on a 3.3 V power supply. The
power consumption results for the sensors and microphone used for the prototype are
shown in Table 3. Apart from the minimum consumption of the temperature and humidity
sensors, the microphone through USB seemed to consume 75 mA on average, no matter
whether it was in standby mode or recording. As it seemed to happen and was proven
by the experiments, the USB port was supplied with power even if the connected device
was not recording. To make the BeeSD device consume less power, a periodic agent was
installed that disabled the USB port when the device was not recording. Under this change,
the consumption of the USB port in standby mode was rapidly reduced to 2 mA.

Table 3. Energy consumption of BeeSD end-node device CPU, peripherals, and attached sensors.

Status Temperature and Temperature Microphone (USB)

Humidity Sensor Sensor Probe
Active 0.5 mA 1.5 mA 82 mA
Standby 100 nA 700 nA 70 mA
75 mA (before disabling
Average 0.2 mA 1 mA 2 mA (after disabling)

The ‘BeeSD’s recorded data, including CSV and WAV files, need to be sent to the cloud
data-logging services for further manipulation, as described in Section 3.2. This process
requires data consumption exceeding 1 GB, depending on the frame rate, when the data are
transmitted as WAV files. Such big data transmissions cannot be transferred over a cellular
LTE/3G/4G network because they would require a significant amount of time to upload,
translating to vast energy expenditure. To solve this issue, BeeSD uses its four-core ARM
CPU to convert WAV files into CSV or JPEG preprocessing FFT footprints and then upload
minimal information to the cloud. In this way, consumption is minimized.

4.3. BeeSD Data-Logging Upload Experiments

The BeeSD system’s directive is to achieve all three KPIs mentioned in Table 2. There-
fore, in each BeeSD sensory probing period, each device records at 22,400 Hz for one minute
every time, as many times as scheduled. As well as this recording time, there is some
time needed for the end-node system to boot up and initialize, as it is powered off after
each probe and put to sleep. Additionally, the greater the WAV raw data size, the more
time is needed to transfer it to the cloud. In this experimental scenario, the Wav file size
was 1.9–2.1 MB for a recording interval of 1 min. This recording time is enough for the
cloud processing service to perform deep learning algorithms on the input delivered. The
analytical results of total transmitted sensory data and total time for the data to be sent to
the cloud via the concentrator over an LTE 3G/4G network are shown in Table 4.
Electronics 2022, 11, 3152 11 of 13

Table 4. Relation between the daily number of probes and the total time in minutes needed for upload
to the BeeSD data-logging services via the concentrator device (using n = 1, 5, and 10 end-node
devices per concentrator).

Number of Total Size Total Time Total Time Total Time

Daily Probes (MB) n = 1 (min) n = 5 (min) N = 10 (min)
5 9.51 8.15 13.32 34.33
12 22.81 16.30 26.64 68.66
24 45.61 39.11 63.92 164.77
48 91.21 78.23 127.84 329.53
64 121.61 104.30 170.45 439.37
128 243.21 208.60 340.89 878.73

Focusing on the ‘BeeSD’s optimal number of daily probes achieving all three KPIs,
the final analysis is given in Table 5, which shows the correlation between the number of
probes and the days of operation using a 10.000 mAh battery. With such a battery, KPI 2
was achieved as the whole system was implemented on the beehive lid, making it easy
to install and use. Table 5 shows the number of daily probes performed compared to the
number of BeeSD end-node devices per concentrator (n). For each value of probes and
devices/concentrator, the total number of operations in days was calculated until the end
node’s battery runs out.

Table 5. BeeSD end-node device energy footprint. The number of daily probes in contrast to the number
of nodes used per concentrator, with bold showing the values that cover the KPI 1 requirement.

Number of Days of Operation Days of Operation Days of Operation

Daily Probes (n = 1) (n = 5) (n = 10)
5 294.51 180.18 69.9
12 147.26 90.1 34.95
24 61.36 37.55 14.56
48 30.68 18.77 7.28
64 23.01 14.08 5.46
128 11.51 7.04 2.73

Using BeeSD simultaneously with 12 daily probe intervals and one device per con-
centrator, the BeeSD device provided a battery life expectancy of more than four months.
In contrast, combining five devices per concentrator reduced battery life expectancy to
3 months. In both ways, KPI 1 was satisfied. Furthermore, if connecting 10 devices per
concentrator, the end nodes’ battery life did not exceed 70 days of operation; thus, the
system can be characterized as marginally low energy. In conclusion, using five beehives
under one concentrator is the optional choice, as it provides BeeSD with a high battery
life expectancy and turns it into a low-cost system (mentioned by KPI 3), as it reduces
telecommunication provider costs. Finally, BeeSD achieved all three Key Performance
Indicators in at least two setup cases (n = 5 and n ≤ 7). This makes BeeSD a promising
low-power, low-cost system.
Beekeepers consider important technology enhancements, IoT devices, and tools
offering conditions monitoring and statistical trends, with notifications for monitoring
condition outliers. Nevertheless, since the cost of a small three-frame beehive with a young
queen is around 60–65 € and that for a single floor 10-frame beehive is around 130–150 €, it
is difficult for beekeepers to purchase technological monitoring devices for each beehive. A
typical example is the limited use of weight scales that require a per-beehive installation
and have prohibitive costs which are close to the cost of a 10-frame beehive (150–200 €). To
these expenses, monthly telecommunication provider costs are also added per beehive.
To overcome the telecommunications costs issue, the BeeSD system uses a central
hive (used by the BeeSD IoT devices; KPI 3). It thus reduces these telecommunication
costs to 1/5 or 1/10 if applied to 5 or 10 beehives. Furthermore, since the one-off cost per
Electronics 2022, 11, 3152 12 of 13

device is estimated to be close to 80–100 € (a similar cost also applies for the concentrator
device), it is still a significant purchase for the small professional beekeeper. However,
if the system is provided as a monthly paid service, the one-off BeeSD equipment costs
can be completely deducted from the beekeepers. Therefore, 30–50 € per 5 or 10 beehives
equipped with the BeeSD IoT device monthly service costs can be considered a low-cost
beekeeping investment.

5. Conclusions
This paper presents a new beehive system capable of logging and visualizing various
beehive events. BeeSD aims to evolve traditional beehives, presenting a new era in the
apiary field of remote sensing. As presented in this paper, BeeSD has been validated in real
conditions. To achieve an efficient solution, the authors defined the key validation indicators
for low-cost, low-energy, and easy installation. From the authors’ experimentation, the
implementation of the IoT BeeSD system attached inside the beehive lid turns it into an
easy-to-install system for every beekeeper.
The BeeSD system’s distributed architecture achieves 80–90% fewer telecommunica-
tion costs for cloud data transmissions while maintaining a low energy footprint, providing
the system with the ability of autonomous operation of at least 69–90 days if 5 or 10 beehives
are utilized per concentrator. Finally, the BeeSD system cloud data-logging and processing
services and user interfaces are capable of real-time data visualization and analysis and can
send real-time push notifications to the beekeepers regarding critical measurements. The
authors will consider data processing and analysis with the incorporation of deep learning
algorithms in future work.

Author Contributions: Conceptualization, D.I.K. and S.K.; Funding acquisition, C.V.B., K.A.S. and
G.S.S.; Investigation, D.I.K.; Methodology, S.K.; Resources, C.V.B., K.A.S. and G.S.S.; Software, D.I.K.
and T.K.; Supervision, S.K.; Writing—original draft, D.I.K.; Writing—review & editing, S.K. All
authors have read and agreed to the published version of the manuscript.
Funding: This research has been co-financed by the European Union and Greek national funds
through the Operational Program Competitiveness, Entrepreneurship and Innovation, under the call
RESEARCH–CREATE–INNOVATE (project code: T21EDK-2402).
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Etxegarai-Legarreta, O.; Sanchez-Famoso, V. The Role of Beekeeping in the Generation of Goods and Services: The Interrelation
between Environmental, Socioeconomic, and Sociocultural Utilities. Agriculture 2022, 12, 551. [CrossRef]
2. Fitzgerald, D.W.; Murphy, F.E.; Wright, W.M.; Whelan, P.M.; Popovici, E.M. Design and development of a smart weighing scale
for beehive monitoring. In Proceedings of the 2015 26th Irish Signals and Systems Conference (ISSC), Carlow, Ireland, 24–25 June
2015; pp. 1–6. [CrossRef]
3. Kviesis, A.; Zacepins, A. System architectures for real-time bee colony temperature monitoring. Procedia Comput. Sci. 2015, 43,
86–94. [CrossRef]
4. Cecchi, S.; Terenzi, A.; Orcioni, S.; Piazza, F. Analysis of the sound emitted by honey bees in a beehive. In Proceedings of the
Audio Engineering Society Convention 147, New York, NY, USA, 16–19 October 2019.
5. Voudiotis, G.; Kontogiannis, S.; Pikridas, C. Proposed Smart Monitoring System for the Detection of Bee Swarming. Inventions
2021, 6, 87. [CrossRef]
6. Milota, M. Beehive Security System against Unauthorized Manipulation. In Proceedings of the 27th Conference EEICT 2021,
Brno, Czech Republic, 27 April 2021.
7. Bellos, C.V.; Fyraridis, A.; Stergios, G.S.; Stefanou, K.A.; Kontogiannis, S. A Quality and disease control system for beekeeping. In
Proceedings of the 2021 6th South-East Europe Design Automation, Computer Engineering, Computer Networks and Social
Media Conference (SEEDA-CECNSM), Preveza, Greece, 24–26 September 2021; pp. 1–4. [CrossRef]
8. Meikle, W.; Holst, N. Application of continuous monitoring of honeybee colonies. Apidologie 2015, 46, 10–22. [CrossRef]
9. Gates, B.N. The Temperature of the Bee Colony; Number 96; US Department of Agriculture: Washington, DC, USA, 1914.
10. Jiang, J.A.; Wang, C.H.; Chen, C.H.; Liao, M.S.; Su, Y.L.; Chen, W.S.; Huang, C.P.; Yang, E.C.; Chuang, C.L. A WSN-based
automatic monitoring system for the foraging behavior of honey bees and environmental factors of beehives. Comput. Electron.
Agric. 2016, 123, 304–318. [CrossRef]
Electronics 2022, 11, 3152 13 of 13

11. Murphy, F.E.; Popovici, E.; Whelan, P.; Magno, M. Development of an heterogeneous wireless sensor network for instrumentation
and analysis of beehives. In Proceedings of the 2015 IEEE International Instrumentation and Measurement Technology Conference
(I2MTC) Proceedings, Pisa, Italy, 11–14 May 2015; pp. 346–351. [CrossRef]
12. Zabasta, A.; Zhiravetska, A.; Kunicina, N.; Kondratjevs, K. Technical Implementation of IoT Concept for Bee Colony Monitoring.
In Proceedings of the 2019 8th Mediterranean Conference on Embedded Computing (MECO), Budva, Montenegro, 10–14 June
2019; pp. 1–4. [CrossRef]
13. Zacepins, A.; Kviesis, A.; Pecka, A.; Osadcuks, V. Development of internet of things concept for precision beekeeping. In
Proceedings of the 2017 18th International Carpathian Control Conference (ICCC), Sinaia, Romania, 28–31 May 2017; pp. 23–27.
[CrossRef]
14. Gil-Lebrero, S.; Quiles-Latorre, F.J.; Ortiz-López, M.; Sánchez-Ruiz, V.; Gámiz-López, V.; Luna-Rodrguez, J.J. Honey bee colonies
remote monitoring system. Sensors 2016, 17, 55. [CrossRef]
15. Mrozek, D.; Gȯrny, R.; Wachowicz, A.; Małysiak-Mrozek, B. Edge-Based Detection of Varroosis in Beehives with IoT Devices with
Embedded and TPU-Accelerated Machine Learning. Appl. Sci. 2021, 11, 11078. [CrossRef]
16. Bilik, S.; Kratochvila, L.; Ligocki, A.; Bostik, O.; Zemcik, T.; Hybl, M.; Horak, K.; Zalud, L. Visual Diagnosis of the Varroa
Destructor Parasitic Mite in Honeybees Using Object Detector Techniques. Sensors 2021, 21, 2764. [CrossRef]
17. Cecchi, S.; Spinsante, S.; Terenzi, A.; Orcioni, S. A smart sensor-based measurement system for advanced bee hive monitoring.
Sensors 2020, 20, 2726. [CrossRef] [PubMed]
18. Zgank, A. IoT-Based Bee Swarm Activity Acoustic Classification Using Deep Neural Networks. Sensors 2021, 21, 676. [CrossRef]
[PubMed]
19. Imoize, A.L.; Odeyemi, S.D.; Adebisi, J.A. Development of a Low-Cost Wireless Bee-Hive Temperature and Sound Monitoring
System. Indones. J. Electr. Eng. Inform. 2020, 8, 476–485. [CrossRef]
20. Ngo, T.N.; Wu, K.C.; Yang, E.C.; Lin, T.T. A real-time imaging system for multiple honey bee tracking and activity monitoring.
Comput. Electron. Agric. 2019, 163, 104841. [CrossRef]
21. Marchal, P.; Buatois, A.; Kraus, S.; Klein, S.; Gomez-Moracho, T.; Lihoreau, M. Automated monitoring of bee behaviour using
connected hives: Towards a computational apidology. Apidologie 2020, 51, 356–368. [CrossRef]
22. Kontogiannis, S. An internet of things-based low-power integrated beekeeping safety and conditions monitoring system.
Inventions 2019, 4, 52. [CrossRef]
23. Kviesis, A.; Zacepins, A.; Durgun, M.; Tekin, S. Application of wireless sensor networks in precision apiculture. In Proceedings of
the 14th International Scientific Conference Engineering for Rural Development, Jelgava, Latvia, 20–22 May 2015; Volume 20,
pp. 440–445.
24. Zygi Company. Digital Scales for Measuring Beehive Weight and Respond to Critical Incidents, Product. 2017. Available online:
http://zygi.gr (accessed on 19 May 2020).
25. 3Bee Company. Hive-Tech, the Innovative Monitoring System for Beekeeping. 2017. Available online: https://www.3bee.it/en/
hive-tech (accessed on 27 September 2022).
26. Arnia Company. Remote Hive Monitoring System with Alerts. 2018. Available online: http://www.arnia.co.uk (accessed on 27
September 2022).
27. Tomtsis, D.; Kokkonis, G.; Kontogiannis, S. Evaluating existing wireless technologies for IoT data transferring. In Proceedings of
the 2017 South Eastern European Design Automation, Computer Engineering, Computer Networks and Social Media Conference
(SEEDA-CECNSM), Kastoria, Greece, 23–25 September 2017; pp. 1–4. [CrossRef]