FROST RISK, IRRIGATION AND FERTILIZATION OPTIMIZATION
WITH ARTIFICIAL INTELLIGENCE SUPPORTED AGRICULTURAL
SYSTEMS
Köksal GÜNDOĞDU 1 Ali ÇALHAN 2
schola @koksalgundogdu.com [email protected].
ORCID Kimliği: 0000-0001-7694-1841 ORCID Kimliği: 0000-0002-5798-3103
Mu aza CİCİOĞLU 3
mu [email protected].
ORCID Kimliği: 0000-0002-5657-7402
1 Elec ical, Elec onic and Compu e Enginee ing, Düzce Uni e si y, Düzce, TÜRKİYE
2 Compu e Enginee ing, Düzce Uni e si y, Düzce, TÜRKİYE
3 Compu e Enginee ing, Bu sa Uludağ Uni e si y, Bu sa, TÜRKİYE
ABSTRACT
The e e -inc easing wo ld popula ion, ine icien use o limi ed esou ces, and he ad e se
e ec s o changing li ing condi ions cause he need o ood o inc ease daily. T adi ional
ag icul u al me hods canno p o ide su icien yields unde di e en en i onmen al condi ions
and land s uc u es, which e eals he need o mo e e ec i e app oaches in ag icul u e. Wi h
he de elopmen o senso echnologies, sa elli e sys ems, and a i icial in elligence-suppo ed
so wa e, p ecision ag icul u e applica ions ha e gained signi ican momen um in ecen
yea s. This s udy p esen s he design o an elec onic ca d de eloped o op imize he p ocesses
o e iliza ion and i iga ion, and p o ec agains he isk o ag icul u al os in ag icul u al
a eas. The designed sys em collec s da a such as soil mois u e, empe a u e, a mosphe ic
p essu e, ambien empe a u e, and humidi y, analyzes hem wi h an a i icial in elligence
algo i hm, and acco dingly p o ides os p e en ion, wa e , and e ilize op imiza ion.
The p oposed sys em p edic s os in ad ance and p e en s he nega i e e ec s o os by
ac i a ing ogging and hea ing mechanisms. I also pe o ms humidi y op imiza ion o p e en
nega i e e ec s such as mic oo ganism sp ead, decay, pho osyn hesis educ ion, and pes
inc ease due o ai humidi y in he ag icul u al land. In his con ex , he s udy o e s an
in eg a ed p ecision ag icul u e solu ion suppo ed by a i icial in elligence o inc ease
ag icul u al p oduc i i y and ensu e he sus ainabili y o na u al esou ces.
Key Wo ds: A i icial in elligence in ag icul u e, p ecision ag icul u e p ac ice, decision ee
algo i hm, sma ag icul u e
I. INTRODUCTION
The apid inc ease in he wo ld popula ion, he inadequa e use o ou esou ces, and he
nega i e e ec s o he changing li e s uc u e a e among he main p oblems ha humani y
mus sol e. These p oblems lead o an inc ease in he need o ood. When we canno use ou
esou ces e ec i ely, mankind will ace he isk o s a a ion soon. Wi h he de elopmen o
echnology and especially a i icial in elligence in ecen yea s, i is possible o op imize ood
p oduc ion. Many o ganiza ions and go e nmen s ha e ecen ly ca ied ou se ious s udies in
his ield. Today, al hough adi ional ag icul u al sys ems and echnologies con ibu e o ood
p oduc ion, hei con ibu ion o p oduc i i y is less han expec ed due o he s uc u e o he
land o he di e ences in c ops in ag icul u al a eas. To inc ease p oduc i i y in ag icul u e,
he concep o P ecision Ag icul u e was i s discussed in he la e 1980s in he USA, Canada,
and some Eu opean coun ies, bu i emained only as a concep due o he insu icien
de elopmen o echnology in hose yea s [1-3]. Wi h he de elopmen o senso s, sa elli e
echnology, da a p ocessing capaci ies in he mid-2000s, and he de elopmen o d ones,
sma phones, and a i icial in elligence-suppo ed so wa e a e 2010, p ecision ag icul u e
ceased o be a concep and s a ed o u n in o p ojec s used in he ield o ag icul u e [4-6].
The ac ha A i icial In elligence da a p ocessing capaci y eached i s peak a e 2020,
compa ed o p e ious yea s, has inc eased he ole and e ec i eness o A i icial In elligence
in P ecision Ag icul u e applica ions [1]. Va ious s udies ha e been ca ied ou o elimina e
he mos basic p oblems in ag icul u e wi h echnology. I has been obse ed ha he PH and
mine al s uc u e o he soil change due o unnecessa y e iliza ion [7, 8]. Due o unnecessa y
i iga ion, p oblems such as wa e loss up o 50%, ene gy consump ion, o ing o plan oo s,
and ungal g ow h ha e been encoun e ed [9, 10]. In addi ion, c op losses ha e occu ed due
o inco ec de ec ion o ag icul u al os [11-13]. All hese a e he main ac o s ha educe
yields in ag icul u e. In addi ion, he sp ead o mic oo ganisms due o ai humidi y, dec ease
in pho osyn hesis, excessi e wa e in ake by he plan , o and mold, lack o oxygen, and
inc ease in pes s due o humidi y cause c op losses in ag icul u al lands [14,15]. I has been
obse ed in he li e a u e ha he e a e a ious p ecision ag icul u e and a i icial in elligence
applica ions o op imize i iga ion, e iliza ion, and ag icul u al os e en s. These
applica ions aim o op imize i iga ion and e iliza ion and o p edic ag icul u al os . In
hese applica ions, he applica ion o i iga ion wi h iming is highly p e e ed, while o
e iliza ion and ag icul u al os , e ilize applica ion by in o ming a me s o sys ems o
comba ag icul u al os is widely used [16-25]. Depending on he condi ion o he land,
complex en i onmen al condi ions, and changing plan species, measu emen esul s a e
e alua ed independen ly o he land. In addi ion, a ious de iciencies a e encoun e ed
depending on he esul s, such as ob aining ag icul u al land in o ma ion om sa elli e [12,
13]. I has been obse ed ha he mois u e and a mosphe ic p essu e pa ame e s on he land
a e no aken in o accoun while p oducing solu ions.
In his s udy, an elec onic ca d design has been ealized o he acquisi ion o p ecision
ag icul u al measu emen esul s and hei e alua ion wi h an a i icial in elligence algo i hm.
Wi h his ca d, a mosphe ic p essu e, soil mois u e, soil empe a u e, ambien humidi y, and
ambien empe a u e da a a e aken om he ag icul u al land as inpu . Fe ilize and wa e
op imiza ion is p o ided acco ding o he ambien condi ions, wi h he wa e ank and
e ilize ank, and e iliza ion and i iga ion a e made acco ding o he needs o he land. In
addi ion, he hea ing and ogging engine connec ed o he sys em helps p o ec he land om
os . The sys em uses an a i icial in elligence algo i hm o igge ogging and hea ing
ac ions be o e and du ing os condi ions, based on measu emen esul s. The humidi y
op imiza ion is ca ied ou o p e en he sp ead o mic oo ganisms caused by ai humidi y,
dec ease in pho osyn hesis, excessi e wa e in ake by he plan , o and mold, lack o oxygen,
and inc ease in ha m ul insec s due o humidi y.
II.SYSTEM DESIGN
A. SYSTEM OVERVIEW
As can be seen in Figu e 1, he sys em is designed in h ee modules. The i s pa is he
“Inpu ” module, whe e he measu emen esul s a e aken om he ag icul u al land. The
second pa is he “Da a P ocessing” module, whe e he measu emen esul s in he Inpu
sec ion a e e alua ed wi h an a i icial in elligence algo i hm, and esul s a e gene a ed. The
hi d pa is he “Ou pu ” module, whe e he esul s p oduced in he Da a P ocessing a e pu
in o p ac ice.
Figu e 1. Block Diag am o he Designed Sys em
The Inpu Module con ains he a mosphe e senso , ambien humidi y senso , ambien
empe a u e senso , soil humidi y senso , and soil empe a u e senso . The Da a P ocessing
Module uns he decision ee algo i hm, which is one o he a i icial in elligence algo i hms,
and p oduces esul s.
The Ou pu Module is whe e he in o ma ion om he Inpu Module is p ocessed by he Da a
P ocessing Module and hen pu in o ac ion. I includes a e ilize and wa e ank, which is
con olled by he Da a P ocessing Module o op imize e iliza ion and i iga ion. I also
ea u es a sys em wi h a hea e , humidi ie , and ogging uni o op imize p o ec ion agains
ag icul u al os and main ain p ope ambien humidi y le els.
B. INPUT UNIT
The Inpu Uni con ains an a mosphe e senso , an ambien humidi y senso , an ambien
empe a u e senso , a soil humidi y senso , and a soil empe a u e senso . A mosphe e senso
is a senso ha can measu e om -500m dep h o 9000m al i ude, acco ding o sea le el. This
senso can ope a e be ween -40°C and +85°C empe a u e le els. Thanks o he senso , he
a mosphe ic p essu e o he ag icul u al land is measu ed. The soil empe a u e senso can
measu e soil empe a u e be ween -55°C and +125°C. I wo ks acco ding o ±0.5°C deg ee o
accu acy. The soil mois u e senso ope a es wi h 10-bi sensi i i y. I has an ope a ing cu en
o app oxima ely 20mA. I wo ks acco ding o he p inciple o elec ical conduc i i y
be ween wo p obes plan ed in he soil. I he soil is mois , conduc i i y inc eases; i he soil
is d y, conduc i i y dec eases. The ambien empe a u e senso can measu e ambien
empe a u e be ween -40°C and +80°C. I wo ks acco ding o ±0.5°C accu acy. The ambien
humidi y senso wo ks wi hin he ange o 0% ~ 100% RH. I has ±1% RH humidi y
sensi i i y.
C. DATA PROCESSING UNIT
In he Da a P ocessing Uni , as can be seen in Figu e 2, he e is an ATMEGA32U4
mic ocon olle . The e is a Mini USB on he ca d o p og amming, and special socke inpu s
o he senso s. I has a Buzze and SMD LEDs on i o gi e a wa ning. The mic ocon olle
wo ks as an 8-bi . I has a 16MHz ope a ing equency, 32KB o lash memo y, 2.5KB
SRAM, 1KB EEPROM, 7 PWM ou pu s, 26 digi al I/O, UART, I2C (TWI), and SPI
communica ion in e aces [26]. The decision-making mechanism uses he decision ee
s uc u e, which is one o he a i icial in elligence algo i hms.
Figu e 2. The Designed Elec onic Boa d
D. OUTPUT UNIT
In he Ou pu Uni , he e a e e iliza ion and wa e anks. The e a e also elec onic al es o
con ol hem. As a esul o soil analysis, he e ilize s ha should be gi en o he ag icul u al
land a e added o he e ilizing ank in dilu ed o m. The sys em is con olled by he
mic ocon olle inside, and he liquid e ilize s a e gi en o he d ip i iga ion sys em in a
con olled manne . In his way, e iliza ion op imiza ion o he soil is ensu ed. Acco ding o
he mois u e con en o he soil and he decision ee algo i hm inside he sys em, he wa e
ank al es a e opened in a con olled manne by he mic ocon olle , and i iga ion
op imiza ion is achie ed h ough he d ip i iga ion sys em.
Figu e 3. The Designed Fan Mo o Sys em
As can be seen in Figu e 3, he e is a an sys em ha includes a hea e , humidi ica ion, and
ogging uni o p e en ag icul u al os e en s and o op imize ambien humidi y. In his
way, acco ding o he esul s o he da a e alua ed in he da a p ocessing uni , he necessa y
hea ing, humidi ica ion, and ogging uni s o inc ease/dec ease he humidi y o o p e en
ag icul u al os a e ope a ed in a con olled manne and sen o he sys em wi h he help o
he an mo o . In his way, ag icul u al os o humidi y op imiza ion is p o ided.
III. FINDINGS AND DISCUSSION
The es sys em in Figu e 4 was ealized o es he sys em shown in Figu e 1 in a eal
en i onmen . A e a ious ials we e pe o med on his sys em, he ope a ion o he sys em
was obse ed, and he esul s we e ob ained. Machine lea ning, a sub-b anch o a i icial
in elligence, was used in ou sys em. Decision ee algo i hm is p e e ed among machine
lea ning algo i hms. Since ou sys em has an 8-bi mic ocon olle , he decision ee algo i hm
makes e y as p edic ions in eal- ime applica ions ou side he aining p ocess on hese
p ocesso s. S udies on he pe o mance and op imiza ion o decision ees in embedded
sys ems show ha ees wi h a dep h o 12 can un wi h la encies less han 50 ns [38]. In
ag icul u e, he p ocess o imp o emen s and in e en ions o minimize nega i e impac s is
called op imiza ion[39]. Fo his eason, we p e e ed o use op imiza ion i les in he con en .
Figu e 4. Physical Fo m o he Sys em
A. AGRICULTURAL FROST OPTIMIZATION
In he li e a u e, soil empe a u e and ambien empe a u e pa ame e s a e mainly used in
s udies on ag icul u al os . I has been obse ed ha soil mois u e pa ame e , ambien
humidi y pa ame e , and especially a mosphe ic p essu e pa ame e , due o soil i iga ion and
o he easons, ha e no been used [16-25]. In ou s udy, i is seen ha e ec i e esul s a e
ob ained when we measu e hese pa ame e s and subjec hem o a decision ee s uc u e.
Decision ees a e widely used in classi ica ion and p edic ion p oblems, especially in da a
mining and machine lea ning. This me hod b anches he da a acco ding o ce ain
cha ac e is ics and p esen s a class label o p edic ion alue a each lea node. In his way,
e ec i e analysis can be pe o med on la ge da a se s and meaning ul esul s can be ob ained
[27].
𝑓(𝑥)=𝑦,𝑖𝑓 𝑥≤ 𝑡∧ 𝑥> 𝑡
𝑦,𝑖𝑓 𝑥≤ 𝑡∧ 𝑥≤ 𝑡
𝑦,𝑖𝑓 𝑥> 𝑡 (1)
In equa ion (1):
x=(x1,x2,...,xn) : Inpu ec o (e.g. senso da a)
xi: i. p ope y (e.g x1= soil mois u e, x2 = ai empe a u e)
i: i. node h eshold alue
yj: class/ac ion (e.g., “Tu n on Fan”, “Tu n on Fog Mo o ”)
A mosphe ic p essu e is also an impo an concep in ag icul u al os . The eezing poin o
wa e is accep ed as 0°C a 1 a m p essu e. As he a mosphe ic p essu e changes, he eezing
poin changes. This change is explained by he Clausius-Clapey on Equa ion [28].
∆
∆ (2)
In equa ion (2):
: De i a i e o eezing poin change o p essu e
ΔV: Volume change du ing phase ansi ion
ΔH : Mel ing ( usion) en halpy
T: Absolu e empe a u e (Kel in)
Acco ding o his equa ion, when he p essu e dec eases, he o ma ion empe a u e ( eezing
poin ) o ice dec eases.
Figu e 5. Phase diag am o pu e wa e be ween p essu e and empe a u e
The ela ionship be ween a mosphe ic p essu e and eezing empe a u e can be seen in he
phase diag am be ween p essu e and empe a u e o pu e wa e , shown in Figu e 5. As can be
seen in Equa ion (2) and Figu e 5, he o ma ion empe a u e o ice dec eases when he
p essu e dec eases.
𝑓(𝑥)=
⎩
⎪
⎪
⎪
⎨
⎪
⎪
⎪
⎧
𝑀,𝑖𝑓 𝑇≤0
𝑀 ,𝑖𝑓 𝑇≤0
𝑀,𝑖𝑓 ∆𝑃<0
𝑀,𝑖𝑓 𝑅𝐻≥80% ∧
<0
𝑀,𝑖𝑓 𝑇→0 ∧ |𝑇−𝑇 |≫0
𝑀 ,𝑖𝑓 𝑇→0 ∧ |𝑇−𝑇 |≫0
𝑀,𝑖𝑓 𝑇𝑙𝑜𝑤∧ 𝑅𝐻 ℎ𝑖𝑔ℎ
𝑁𝑜 𝐴𝑐𝑡𝑖𝑜𝑛, 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
(3)
In equa ion (3):
𝑇: Soil empe a u e (°C)
𝑇: Ai empe a u e (°C)
𝑅𝐻: Ai humidi y (% ela i e humidi y)
𝑃: A mosphe ic p essu e
∆𝑃: Sudden change in a mosphe ic p essu e (de i a i e)
𝑀: Hea e mo o ac i e
𝑀 : Fogging mo o ac i e
The decision ee s uc u e in Equa ion (3) is execu ed in he da a p ocessing uni . In his way,
ag icul u al os p e en ion op imiza ion wo ks success ully by using soil empe a u e, ai
empe a u e, a mosphe ic p essu e, and ai humidi y da a ob ained om he inpu uni .
B. AIR HUMIDITY OPTIMIZATION
In he li e a u e, i is seen ha he sp ead o mic oo ganisms, a dec ease in pho osyn hesis,
excessi e wa e in ake by he plan , o and mold, lack o oxygen, and an inc ease in pes s due
o humidi y caused by ai humidi y [29]. In o he wo ds, ambien humidi y in ag icul u e is
also an impo an pa ame e ha con ibu es o p oduc i i y. To p e en he nega i e e ec s o
humidi y, humidi y op imiza ion is ca ied ou . The measu emen esul s a e applied o he
decision ee algo i hm as shown in equa ion (1). The decision ee algo i hm o he applied
da a is shown in equa ion (4).
𝑓(𝑅𝐻,𝑇 )=𝑀,𝑖𝑓 𝑅𝐻≥80% ∧ 𝑇<5
𝑀 ,𝑖𝑓 𝑅𝐻≥80% ∧ 𝑇≤0
𝑀 ,𝑖𝑓 𝑅𝐻≤40%
𝑁𝑜 𝐴𝑐𝑡𝑖𝑜𝑛, 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒 (4)
𝑇: Ai empe a u e (°C)
𝑅𝐻: Ai humidi y (% ela i e humidi y)
𝑓 (𝑅𝐻,𝑇 ): Decision unc ion ha de e mines ac ion based on humidi y
and empe a u e inpu s
𝑀: Hea e mo o ac i e
𝑀 : Fogging mo o ac i e
𝑀 : Humidi ica ion mo o ac i e
The ambien humidi y op imiza ion is achie ed by unning he decision ee algo i hm in
equa ion (4) in he da a p ocessing uni . In o he wo ds, he humidi y alue is kep unde
con ol wi hin he desi ed alue ange. The RHai alue he e may a y depending on he
ag icul u al p oduc o be g own. In o ma ion abou his is gi en in Table 1[30].
Table 1. Recommended Ambien (Ai ) Humidi y Values in Ag icul u al P oduc ion
Plan Type / En i onmen
Recommended Rela i e
Humidi y (%)
Explana ion
Vege ables
( oma oes, peppe s) 60–80
High humidi y can inc ease
ungal diseases, so cau ion
should be exe cised.
F ui ees 50–70
Mois u e con ol is impo an
du ing he lowe ing pe iod.
Ce eals (whea , ba ley) 40–60
Low humidi y educes he isk
o disease.
G eenhouse en i onmen 70–85
High humidi y may be equi ed
in a con olled en i onmen .
C. WATER OPTIMIZATION
O e -i iga ion leads o ine icien use o esou ces. I also educes p oduc i i y. While 50%
wa e sa ing is achie ed in con olled i iga ion, i leads o e icien and economical use o
ene gy esou ces. While unnecessa y i iga ion dis up s he PH and mine al balance o he
soil, i also leads o oo o ing o plan s [31, 32]. I iga ion acco ding o soil mois u e le el is
an e ec i e i iga ion me hod. Recommended soil mois u e anges in ag icul u al lands a e
shown in Table 2[33, 34].
Table 2. Recommended Soil Mois u e Ra ios in Ag icul u al Lands
Plan Type
Recommended Soil Mois u e
Range (%) Explana ion
Vege ables
( oma oes, peppe s, e c.) 60–80%
Cons an ly mois , bu no
puddles
F ui ees
50
–
70%
Flowe ing and ui ing c i ical
Ce eals (whea , ba ley) 40–60%
Impo an du ing ge mina ion
and sp ou ing
Legumes
40
–
60%
Excess wa e can cause oo o
G eenhouse plan s 70–85%
Requi es
high humidi y in
con olled en i onmen
The measu emen esul s a e applied o he decision ee algo i hm as shown in equa ion (1).
The decision ee algo i hm o he applied da a is shown in equa ion (5).
𝑓(θ)= 𝑀,𝑖𝑓 θ<θ
𝑀 ,𝑖𝑓 θ>θ
𝑁𝑜 𝐴𝑐𝑡𝑖𝑜𝑛,𝑖𝑓 θ≤θ≤θ (5)
θ: Soil mois u e con en (%)
θ: Minimum mois u e h eshold (Fo example, %30)
θ: Maximum humidi y h eshold (Fo example, %70)
𝑓(θ): Decision unc ion
𝑀: I iga ion engine in ope a ion
No Ac ion: No ac ion is aken
𝑀: I iga ion is s opped
Soil mois u e op imiza ion is achie ed by unning he decision ee algo i hm in equa ion (5)
in he da a p ocessing uni . In o he wo ds, he soil mois u e alue is con olled wi hin he
desi ed alue ange.
