MICROPROCESSOR AND INTERFACING – CSE2006

J – COMPONENT

SMART IRRIGATION USING ARTIFICIAL INTELLIGENCE AND IOT

SLOT – L27+L28/ C2

FACULTY – PROF. BHULAKSHMI BONTHU

GROUP MEMBERS:
NAMAN AGRAWAL 18BCE0107
RITIK GUPTA 18BCE0154
AYUSH SHARMA 18BCE0172
WORK DONE BY ALL TEAM MEMBERS:
1. AYUSH SHARMA (18BCE0172)
 IOT Device build up, Arduino coding and its integration with android app
 Coded PLSR algorithm in IOT device
 Connected IOT device to cloud using Arduino IDE coding and connected it to Android app
too
2. RITIK GUPTA ( 18BCE0154)
 Coded backend API to communicate with Historic data
 Coded API to fetch important factors for moisture content calculation.
 Worked on PLSR Algorithm in integrating it with backend API

3. NAMAN AGRAWAL (18BCE0107)

 Built Android App and connected it with IOT device
 Coded backend for cloud server (Server side code)
 Connected app and cloud together with IOT device. (Coded in Android Studio)

SMART IRRIGATION IOT DEVICE AND ANDROID APP DEMO VIDEO (FULL
PROJECT DEMO VIDEO):
 GOOGLE DRIVE LINK:
https://drive.google.com/file/d/1by94dBEA6mBmVAaIhstissiM-
osbbpRh/view?usp=sharing
 YOUTUBE LINK:
https://youtu.be/j_sHtFuISoo

ANDROID APP DOWNLOAD LINK (APK):

 GOOGLE DRIVE LINK (YOU CAN DOWNLOAD THE APP SHOWN IN VIDEO
FROM HERE) :
https://drive.google.com/file/d/1Nb7frR7sRSwNL9whgJMeKEHGU_frI_gN/view?us
p=sharing
 BONAFIDE CERTIFICATE

This is to certify that the project entitled “SMART IRRIGATION USING ARTIFICIAL
INTELLIGENCE AND IOT” is the bonafide work of Ayush Sharma(18BCE0172), Ritik
Gupta(18BCE0154) and Naman Agrawal(18BCE0107) done in the fourth semester of the
academic year 2019-2020. He has duly completed his project and has fulfilled all the requirements
of the course CSE2006, J component, to my satisfaction.

Ms. BHULAKSHMI BONTHU

SUPERVISOR,
Assistant Professor (Senior),
VIT, Vellore.

Date:
ABSTRACT
The farm can be modernized with electronic technology that continuously monitors the conditions of plants and
soil, so that the plants could be provided watering as required. All of this can be controlled and monitored online
with the application of IoT. We will make a device will be interfaced with a water pump controlled
by L293D motor driver IC. The water pump is automatically controlled based on the values of the various
environmental factors like temperature, humidity, soil moisture and light intensity which we can measure
through sensors like DHT-11 Temperature and Humidity sensor, Moisture sensor and LDR sensor. Through our
research work we will try to make the model intelligent by storing the previous scanned values in database and
doing the pre-perception on the basis of historic values stored in learning phase of the working model. We are
using PLSR algorithm to train our model.
Agriculture plays a consequential role in the economy and its contribution is predicated on quantifiable crop yield
which is highly dependent upon irrigation. In a country like India, where agriculture is largely predicated on the
unorganized sector, irrigation techniques and patterns followed are inefficient and often lead to nonessential
wastage of water. This calls for the desideratum of a system which can provide an efficient and deployable
solution. In this paper, we provide an Automatic Irrigation System predicated on Artificial Perspicacity and
Internet of Things, which can autonomously irrigate fields utilizing soil moisture data. The system is predicated
on prognostication algorithms which make utilization of historic weather data to identify and presage rainfall
patterns and climate changes; thereby engendering an astute system which irrigates the crop fields selectively
only when required as per the weather and authentic-time soil moisture conditions. The system has been tested in
a controlled environment with an 80 percent precision, thus providing an efficient solution to the quandary.

INTRODUCTION
India follows traditional agricultural methods in irrigation practices [1]. Irrigation is a paramount factor in
determining the crop yield and largely varies with the geographical, climatic, and topological factors. Farmers
primarily depend on personal monitoring and their experience in irrigating the fields, and as a result, irrigation
becomes largely inefficient and eccentric. India, ergo requires a simple irrigation solution on which the farmers
can depend indefinitely, which can habituate to the local climatic conditions, and accurately soothsay the quantity
of water required by the crops in authentic time to ascertain judicious utilization of water resources, and
additionally a better crop yield. The main concern in India is not water shortage but water wastage, and poor
utilization of the resources due to lack of vigilance, facilities, and infrastructure. Due to wastage of water, the
country has already suffered through immense drought conditions, varied rainfall patterns, and immensely
colossal economic losses due to the eradication of crops [2].

Traditional automatic irrigation systems are not opportune for India, as they cannot habituate to the transmuting
rainfall patterns and do not respond well to geographical changes. Thus, we have developed a perspicacious
system which can study the patterns of rainfall in a region, and soothsay weather conditions in order to habituate
to the geography, thus prognosticate the quantity of water needed for irrigation, to minimize wastage and
increment the crop yield [3]. To achieve this, we make utilization of Node MCUs, and soil moisture sensors,
placed inside waterproof boxes and spread evenly throughout the area to be irrigated. All these nodes are
connected to cloud via Wi-Fi connection [4]. The system analyses soil moisture content through the deployed
sensors, which is utilized to soothsay the quantity of water required to irrigate the area with an appreciable
precision. It is a highly autonomous system which requires little to no human intervention once deployed in a
field. The system developed makes utilization of Desultory Forest Regressor to soothsay the weather [5]. It makes
utilization of the traditional data of rainfall patterns and weather data. It has the faculty to gradually
acclimate to the region-concrete climatic conditions and its precision amends with every presage made.
The system is designed to be updated every 30 minutes as a diminutive customary interval which makes it
power efficient as well. After every such conventional interval, it updates the dataset as per the incipient
sensor data provided. The system switches on the motor or pump if the soil moisture content detected is deficient
and predicated on the sensor data we make utilization of Partial Least Square Regression (PLSR) algorithm
to presage the quantity of water required [6]. With this, we are able to calculate the time interval for which the
motor should pump water. Additionally, in case the system detects adequate soil moisture or rain is presaged
then the pump is not switched on while the sensor measures the soil moisture capacity after the
precipitation. The system designed is power efficient, water efficient and low on maintenance. The systems are
scattered throughout the area of the farm. Thus, we can switch on the drip or sprinkler for a particular area rather
the entire farm in order to increment efficiency [7]. This avails in minimization of water wastage, a better
understanding of crop water capacity and patterns required for efficient irrigation. Withal, the nodes work
on a replication predicated system so it makes identification of any malfunction more facile. The health of the
nodes can be monitored through a mobile app predicated on the mapping of the farm and area designated for the
irrigation. Thus, the system promotes low maintenance and proves to be effective.

LITERATURE REVIEW
Before going into the details of our Intelligent IoT based Automated Irrigation system, we will review some of
the existing system related to Agriculture.

[1] INTELLIGENT IRRIGATION SYSTEM USING ARTIFICIAL INTELLIGENCE AND MACHINE

LEARNING: A COMPREHENSIVE REVIEW (September 2018)
RESEARCH PAPER KEY POINTS:

 Temperature and moisture sensors will sense the soil condition at regular intervals and send the data to
the microcontroller of the Arduino. Arduino has an inbuilt analog to digital converters which will convert
input analog data to a digital one.
 Machine learning based KNN algorithm is employed in the hardware. This algorithm is very simple and
rudimentary. The sensed temperature and moisture values are then engulfed in the machine learning
algorithm and based on the training values; Raspberry pi3 sends the signal to Arduino to start the pump
for irrigation process.
 Regardless of water supplied, every time the data is processed in the algorithm, it also updates the
database. This database holds the information of sensor values and number of irrigation processes.
 The idea is to implement ANN and KNN method both in the same system. Raspberry Pi module has
algorithms of the KNN method which helps the machine to learn from the previous data set of crop
selection, crop growth and crop yield to give the data to the farmer.

DRAWBACK OR LIMITATIONS:

 After reading the above research paper completely, we found a genuine and a big loophole in which the
above proposed system do not involves a rainfall prediction system.
 There can be a situation in which the system watered the plant and just after this process rainfall begins.
This can become a huge disaster for the farmers as over-irrigation can affect the crops very badly.
[2] Smart irrigation system for environmental sustainability in Africa: An Internet of Everything (IoE)
approach (Published: 13 June 2019)
RESEARCH PAPER KEY POINTS:

 The research paper show how the control and monitoring of a solar powered smart irrigation system can
be achieved using sensors and environmental data from an Internet of Everything (IoE).
 The collected data is used to predict environment conditions using the Radial Basis Function Network
(RBFN).
 The predicted values of water level, weather forecast, humidity, temperature and irrigation data are used
to control the irrigation system. A web platform was developed for monitoring and controlling the system
remotely.
 Used fuzzy logic rules to decide on the amount of water needed based on the current moisture in soil,
humidity, and time of the day. Summarizing, the need for optimization of water usage is recognized by all
state-of-the-art irrigation systems, and the authors employ a variety of solutions based on sensor networks,
microcontrollers and machine learning (or fuzzy logic) to evaluate and predict water needs for irrigation
in order to ensure efficient use of water.

DRAWBACK OR LIMITATIONS:

 After reading the above research paper completely, we found a genuine and a big loophole in which the
above proposed system do not involves a rainfall prediction system.
 There can be a situation in which the system watered the plant and just after this process rainfall begins.
This can become a huge disaster for the farmers as over-irrigation can affect the crops very badly.
 It will be very difficult for a small scale farmer to use the developed web app. As it will require a desktop
or laptop and computer knowledge.

[3] A Study on Smart Irrigation Using Machine Learning (Published Date: June 10, 2019)
RESEARCH PAPER KEY POINTS:

 Different types of Sensor devices used to measure the environment condition such as soil moisture, soil
temperature, humidity, weather condition, leaf sensing, and canopy temperature. The real time values are
measured by the deployed environmental sensors in irrigation.
 The system was trained through the Levenberg– Marquardt (LM) and back propagation (BP) algorithms
using MATLAB R2007b program. This system achieved 23.9% energy saving and 20.46% water saving
during night-time irrigation and by supplying water to the plant as required.

Traditional Agricultural Monitoring

In some of the traditional irrigation system, irrigation is scheduled by monitoring the soil and water status by
employing tension meter and drip irrigation by automating the controller system in sandy soil.

In some irrigation system, fuzzy logic controller [10] been implemented for an efficient irrigation system for
field having different crops. Fuzzy logic increases the accuracy of measured value and accordingly assists in
decision making. Green house based modern agriculture need to

be precisely controlled in terms of humidity and temperature. The atmospheric conditions of plants inside the
greenhouse vary from place to place which makes it difficult towards maintaining uniformity at all places in
farmhouse manually. So towards this, GSM been used towards reporting the status about irrigation for farmer’s
mobile handset.

ICT Based Agricultural Monitoring

Sensor based paddy growth monitoring system [11] been developed by researchers Kait towards improving
the rice productivity or yield. This system been considered to be cost effective as well as durable at outdoor
operations. The architecture of the system is shown in Figure 1 where sensor nodes like Temperature, humidity,
light, water level etc. are deployed in the field to gather the appropriate data and accordingly transmitted to the
Base station using multi hop routing. The data are processed locally at the BS which are then sent to the remote
server for further processing and analysis. After data analysis, message sent to farmers notifying the field
conditions and providing suggestions.. Knowledge base can be created based on the data collected using Expert
systems for further analysis. This system is not scalable and cannot be reused on other applications as it is not
based on the concept of Machine to Machine Communications (M2M) which cover a large geographical area
with many sensors deployed in the field.

Research [12] been carried out by employing Bluetooth Wireless Transmitter that sense soil moisture,
temperature etc and accordingly send the data to the Base station (BS) which makes the decision towards irrigation
decision based on field and time. The irrigation control unit which is responsible towards irrigating the field
pertaining to operating the sprinkler would receive the control signal from the BS. This is based on water
Requirement of the fields. Some researchers are also working towards Variable rate Sensor based Irrigation
System.

Researchers Wall and King [13] developed an automated field specific irrigation system with soil moisture
sensor and sprinkler valve controller. These systems do not take into consideration monitoring the water pollution
in lakes or rivers and also do not consider M2M Communication concept.

Research been carried out in developing an intensive sensor based irrigation monitoring system which is
scalable and self-organizing [14] The system architecture of their proposed sensor-based irrigation monitoring
system is divided into two layers: bottom layer and upper layer as shown in Figures 2 and 3. Hierarchical sensor
network are placed in bottom layer where nodes are placed in widely separated clusters. These nodes send the
data to Base station (BS) which are connected Wireless LAN that holds the data logger software.[14]. Upper layer
consist of five modules which are “acquisition module, network management module, alarm/network status
display module and business module”. Real and non-real time data are collected using the data gathering modules
from sensor network which are stored in database for decision/alert notifications. Alert notifications or displaying
information to the end users are carried out by the

Alarm/Network status display module. The Alarm/Network status display module acts as an access point between
end users and other modules/networks.

The condition of networks such as localizations, collision and network configuration are carried out by network
management module. This system here developed have introduced the concept of M2M communication where
water and energy conserved by using intelligent sensor network and efficient routing protocol Research also been
done in Crop field assessment towards irrigations, applying fertilizers and pesticides. So accordingly a sensor
network based field management system been developed where solar powered moisture nodes and low resolution
camera deployed. The information from the two sources are captured for comparative assessment and analysis.
Crop height, coverage and greenery information are sent through the sensor networks to the Base Station (BS) by
means of Self-contained, self-powered low resolution camera. In addition crop images are sent from these camera
nodes by allocating the time. Lastly the cattle position and behavior can be observed by means of camera nodes.

Research [15] also been carried out by developing an automated irrigation system (A2S) which is based on sensor
network. Wireless sensors are being employed for monitoring and controlling the agricultural fields. The
management sub system controls the sensor network and accordingly provides service to farmer’s by means of
PDA. In this system, long distance communicated provided by means of WLAN between sensor network and
server. Management subsystems consist of database, application and web server. Data from sensors are received
by the application server which are stored in the database server. Web server is accessed by farmer’s using PDA.
In addition to the above mentioned research employing wireless sensors in agricultural monitoring, mobile phones
also have been adopted in many rural areas by the farmers

[16]. In one of the research, information about the seeds is delivered to the farmer’s in two ways. First is the
periodic broadcast of seed information from web-portal through SMS on mobile phones of farmers depending
upon the season. Second, farmer can query the system by sending names of the crops in SMS and our system will
automatically reply with the available seed information of the crop. So this way farmer will be able to get
information on their mobile phones at a reasonable cost as compared to expenses incurred in travelling to
agricultural offices. Moreover, wastage in time is also an issue that will be solved through this system Low cost
technological solutions are provided by sensor networks for PA towards in field crop management. These crop
conditions/growth are monitored for a longer period of time, remotely make decision and accordingly evaluate
the potential of new crops. With the data collected by sensors, database or knowledge base created. GUIs are
integrated with these sensor networks in these monitoring systems.

Machine Learning in Agricultural Monitoring

The machine learning algorithm has various uses in the field of agriculture monitoring which are being discussed
here.

In one of the research [17], Machine learning been applied towards Grape cultivation. In here farmers are unable
to identify the disease manually on the grapes. The disease on grapes is identifiable only after the infection which
takes lot of time and also has adverse effect on the vineyards. So accordingly a monitoring system developed for
grape cultivation where temperature, relative humidity and leaf wetness sensors are deployed in the vineyard. The
data collected at regular intervals are sent using Zigbee module to the server. The server here employs the hidden
markov model algorithm towards training the data sets pertaining to Temperature, relative humidity and leaf
wetness for analysing the data towards predicting the chance of disease on grapes before getting infected. This
information is sent as alert message via sms to the farmer and expert. The system employs machine learning in
early and accurate detection of disease in grapes and suggests pesticide to protect the crop from disease and reduce
manual disease detection efforts. Also this system can be helpful for farmer’s towards giving information on
schedule of fertilizer’s, pesticide spraying, irrigation etc which would help in improving the quality and quantity
of grapes.

Extreme Machine Learning [8] been employed which is a simple, reliable and efficient single hidden layer feed
forward neural networks. This methodology is based on weather factors and time series of soil moisture. Data
sets are obtained from Dookie Applied Orchard in Victoria, Australia. Results have shown that future trend of
soil moisture can be predicted accurately and accordingly decision support can be developed for irrigation
scheduling. This method provides better accuracy compared to Support Vector machines which is the traditional
and conventional soil moisture forecasting In another
Research [19], Random Forest Machine algorithm been developed for detection and classification of different.
Grape diseases like Anthracnose, Powdery Mildew and Downy Mildew from the images collected under
uncontrolled environment with random background. In here, the performance of Random forest been compared
with Probabilistic Neural Network, Back propagation Neural Network and Support Vector Machine. Also
performance of different texture features been studied. The proposed system achieves best classification accuracy
of 86% using Random Forest and GLCM features for background separation and disease classification.

Crop Selection Method research [20] been developed towards selecting the sequence of crops planted over a
season. Selection of Crops resolved based on prediction yield rate which is dependent on parameters like weather,
soil type, water density, crop type etc. Crop, Sowing time, plantation days and predicted yield rate are given as
input for the season in this method and accordingly sequence of crops given as output. The performance and
accuracy of CSM is dependent on predicted value of influenced parameters.

COMPONENTS AND TECHNOLOGY USED

 NodeMCU (ESP8266 Microcontroller)

NodeMCU is an open source firmware for which open source prototyping board designs are available.
The name "NodeMCU" combines "node" and "MCU" (micro-controller unit). The term "NodeMCU"
strictly speaking refers to the firmware rather than the associated development kits. Both the firmware and
prototyping board designs are open source. The firmware uses the Lua scripting language. The firmware
is based on the eLua project, and built on the Espressif Non-OS SDK for ESP8266. It uses many open
source projects, such as lua-cjson and SPIFFS. Due to resource constraints, users need to select the
modules relevant for their project and build a firmware tailored to their needs. Support for the 32-bit
ESP32 has also been implemented.
The prototyping hardware typically used is a circuit board functioning as a dual in-line package (DIP)
which integrates a USB controller with a smaller surface-mounted board containing the MCU and antenna.
The choice of the DIP format allows for easy prototyping on breadboards. The design was initially was
based on the ESP-12 module of the ESP8266, which is a Wi-Fi SoC integrated with a Tensilica Xtensa
LX106 core, widely used in IoT applications.
  NODEMCU PIN CONFIGURATIONS

Pin Name Description

Category

Power Micro-USB, Micro-USB: NodeMCU can be powered through the USB port
3.3V, GND, Vin
3.3V: Regulated 3.3V can be supplied to this pin to power the board
GND: Ground pins
Vin: External Power Supply

Control EN, RST The pin and the button resets the microcontroller
Pins

Analog A0 Used to measure analog voltage in the range of 0-3.3V

Pin

GPIO Pins GPIO1 to NodeMCU has 16 general purpose input-output pins on its board
GPIO16

SPI Pins SD1, CMD, NodeMCU has four pins available for SPI communication.
SD0, CLK

UART TXD0, RXD0, NodeMCU has two UART interfaces, UART0 (RXD0 & TXD0) and
Pins TXD2, RXD2 UART1 (RXD1 & TXD1). UART1 is used to upload the
firmware/program.

I2C Pins NodeMCU has I2C functionality support but due to the internal
functionality of these pins, you have to find which pin is I2C.
 DHT11 Sensor

The digital temperature and humidity sensor DHT11 is a composite sensor that contains a calibrated digital
signal output of temperature and humidity. The technology of a dedicated digital modules collection and
the temperature and humidity sensing technology are applied to ensure that the product has high reliability
and excellent long-term stability. The sensor includes a resistive sense of wet component and an NTC
temperature measurement device, and is connected with a high-performance 8-bit microcontroller.

 Soil moisture sensor

Soil moisture sensors measure the volumetric water content in soil. Since the direct gravimetric
measurement of free soil moisture requires removing, drying, and weighing of a sample, soil moisture
sensors measure the volumetric water content indirectly by using some other property of the soil, such as
electrical resistance, dielectric constant, or interaction with neutrons, as a proxy for the moisture content.
The relation between the measured property and soil moisture must be calibrated and may vary depending
on environmental factors such as soil type, temperature, or electric conductivity. Reflected microwave
radiation is affected by the soil moisture and is used for remote sensing in hydrology and agriculture.
Portable probe instruments can be used by farmers or gardeners.
Soil moisture sensors typically refer to sensors that estimate volumetric water content. Another class of
sensors measure another property of moisture in soils called water potential; these sensors are usually
referred to as soil water potential sensors and include tensiometers and gypsum blocks.
 1- channel Relay Module

A relay is an electrically operated device. It has a control system and (also called input circuit or input
contactor) and controlled system (also called output circuit or output cont actor). It is frequently used in
automatic control circuit. To put it simply, it is an automatic switch to controlling a high-current circuit
with a low-current signal.

The advantages of a relay lie in its lower inertia of the moving, stability, long-term reliability and small
volume. It is widely adopted in devices of power protection, automation technology, sport, remote control,
reconnaissance and communication, as well as in devices of electro mechanics and power electronics.
Generally speaking, a relay contains an induction part which can reflect input variable like current,
voltage, power, resistance, frequency, temperature, pressure, speed and light etc. It also contains an
actuator module (output) which can energize or de-energize the connection of controlled circuit. There is
an intermediary part between input part and output part that is used to coupling and isolate input current,
as well as actuate the output. When the rated value of input (voltage, current and temperature etc.) is above
the critical value, the controlled output circuit of relay will be energized or de-energized.

 Motor Pump (mini DC 12V)

A DC motor is any of a class of rotary electrical motors that converts direct current electrical energy into
mechanical energy. The most common types rely on the forces produced by magnetic fields. Nearly all
types of DC motors have some internal mechanism, either electromechanical or electronic, to periodically
change the direction of current in part of the motor.
DC motors were the first form of motor widely used, as they could be powered from existing direct-current
lighting power distribution systems. A DC motor's speed can be controlled over a wide range, using either
a variable supply voltage or by changing the strength of current in its field windings. Small DC motors
are used in tools, toys, and appliances. The universal motor can operate on direct current but is a
lightweight brushed motor used for portable power tools and appliances. Larger DC motors are currently
used in propulsion of electric vehicles, elevator and hoists, and in drives for steel rolling mills. The advent
of power electronics has made replacement of DC motors with AC motors possible in many applications.
 Breadboard

A breadboard is a construction base for prototyping of electronics. Originally the word referred to a
literal bread board, a polished piece of wood used for slicing bread. In the 1970s the solderless
breadboard (a.k.a. plugboard, a terminal array board) became available and nowadays the term
"breadboard" is commonly used to refer to these.
Because the solderless breadboard does not require soldering, it is reusable. This makes it easy to use for
creating temporary prototypes and experimenting with circuit design. For this reason, solderless
breadboards are also popular with students and in technological education. Older breadboard types did
not have this property. A stripboard (Veroboard) and similar prototyping printed circuit boards, which
are used to build semi-permanent soldered prototypes or one-offs, cannot easily be reused. A variety of
electronic systems may be prototyped by using breadboards, from small analog and digital circuits to
complete central processing units (CPUs).

 Connecting wires

A jump wire (also known as jumper wire, or jumper) is an electrical wire, or group of them in a cable,
with a connector or pin at each end (or sometimes without them – simply "tinned"), which is normally
used to interconnect the components of a breadboard or other prototype or test circuit, internally or with
other equipment or components, without soldering.
Individual jump wires are fitted by inserting their "end connectors" into the slots provided in a breadboard,
the header connector of a circuit board, or a piece of test equipment.
Jumper wires typically come in three versions: male-to-male, male-to-female and female-to-female. The
difference between each is in the end point of the wire. Male ends have a pin protruding and can plug into
things, while female ends do not and are used to plug things into. Male-to-male jumper wires are the most
common and what you likely will use most often. When connecting two ports on a breadboard, a male-to-
male wire is what you’ll need.
 Google Firebase Cloud

Firebase is a Backend-as-a-Service — BaaS — that started as a YC11 startup and grew up into a next-
generation app-development platform on Google Cloud Platform.
Realtime Database
Real-time data is the way of the future. Nothing compares to it. Most databases require you to make
HTTP calls to get and sync your data. Most databases give you data only when you ask for it. When you
connect your app to Firebase, you’re not connecting through normal HTTP. You’re connecting through
a WebSocket. WebSockets are much, much faster than HTTP. You don’t have to make individual
WebSocket calls, because one socket connection is plenty. All of your data syncs automatically through
that single WebSocket as fast as your client’s network can carry it. Firebase sends you new data as soon
as it’s updated. When your client saves a change to the data, all connected clients receive the updated
data almost instantly.
File Storage
Firebase Storage provides a simple way to save binary files — most often images, but it could be
anything — to Google Cloud Storage directly from the client. Firebase Storage has it’s own system of
security rules to protect your Cloud bucket from the masses, while granting detailed write privileges to
your authenticated clients.
Hosting
Firebase includes an easy-to-use hosting service for all of your static files. It serves them from a global
CDN with HTTP/2.

 ARDUINO IDE

The Arduino Integrated Development Environment (IDE) is a cross-platform application

(for Windows, macOS, Linux) that is written in functions from C and C++ . It is used to write and
upload programs to Arduino compatible boards, but also, with the help of 3rd party cores, other vendor
development boards.
METHODOLOGY
The system designed consists briefly of two major components:

1. Artificial Intelligence to predict the amount of rainfall and the soil moisture content.
2. Implementation of system using Internet of Things.

 Central Controller
A central system is utilized to control the nodes, and to decide the time for which the pump affixed to any
particular node is switched on. This implementation utilizes a Cloud API controller [4]. The controller
utilizes a Python3 and JAVA script to execute all its functions. Support scripts are imported to the main
script, and the control logic calls methods from the fortification scripts as and when required. The
functions include communication with the nodes, data preprocessing and deriving inferences utilizing
PLSR algorithms. Training is performed on a high-end computer, and the trained models are then deployed
in the engenderment environment.

 Distributed Nodes
Each node is comprised of an ESP8266 [8] predicated microcontroller board (NodeMCU) and an array of
analog sensors. These sensors are acclimated to fetch information about weather conditions, which is
converted to digital 10-bit integers utilizing the Analog to Digital converters present on each NodeMCU.
An algorithm converts this data to a 32-bit floating point value required by the inference model, on the
NodeMCU itself. Each node contains adscititious support hardware to provide required electrical power
to the NodeMCU and the sensors and to avail facile debugging in the event of failure of the Node. The
puissance circuit consists of a 9V battery input given to a 3.3V regulator. The 3.3V output is then used to
power the NodeMCU and the sensors [13]. A mundane Vcc line and GND line are habituated to supply
power to the aforementioned components. Adscititiously, the 9V supply is directly provided to an L293D
motor drive IC, which potencies the water pump predicated on the inputs it receives from the NodeMCU.
The implementation may withal include one or more solar panel(s) to power the NodeMCU and the
sensors, utilizing battery power only when the pump is required. The pre-processed data, along with some
adscititious bytes used to identify each individual node, is sent to the central controller via the HTTP GET
request.

 THE NETWORK
The nodes are connected to the central controller through a shared WiFi access point, over the IPv4
protocol [9]. A client-server architecture is utilized for communication between the nodes and the central
controller. The nodes do not communicate with each other. This implementation utilizes a Cloud server
on the central controller. The server elongates a URL in the form of a webhook. The nodes connect to the
server via this webhook to send data. The HTTP GET request is utilized to transmit the sensor data after
post-calculation to the server [14]. The default replication of the server conveys two messages to the node
that sent the request: 1. if the pump annexed to the particular node has to be switched on or not, and 2. the
duration for which the pump has to be switched on. The duration is calculated utilizing the output of the
regression algorithm. Each node sends a request to the server at an interval of 30 minutes. The network is
thus essentially a wireless Local Area Network. Since each node works independent of other nodes,
incidental failure of a node does not result in the collapse of the entire system. Supplementally, isolation
of the failed node becomes facile. The IPv4 Address of the central controller is hard-coded into each
 NodeMCU. The central controller is assigned a static IPv4 address utilizing the access point settings. The
port utilized for communication is set to 8080 for this implementation.

 Rainfall Prediction using Artificial Intelligence

The rainfall presage involved a two-phase solution:
• Prediction of Probability of Rainfall in the next 30 minutes
• Estimation of the Amplitude of Rainfall
The first phase is to avail the network to realize whether there is a chance of rainfall to occur in
the next 30 minutes or not and the contrivance keeps checking the status at customary intervals of time
when it is switched on periodically. This is achieved by soothsaying the probability of rainfall to occur
in the next 30 minutes. The system is made power efficient by not perpetually keeping it on for checking
the status of rainfall rather checks at customary time intervals itself. The amplitude of rainfall depends
on multiple parameters including mean of temperature, pressure, maximum and minimum sultriness as
well as the mean dew associated with the air [15]. The dataset used is accountable for local areas and
regions since it contained parameters which can be generalized to presage the rainfall and had desultory
values of all the parameters. The data used is taken from the rainfall data available on the Regime of
India Portal for local regions. The initial deployment of the model took account of the data provided for
the state of Punjab primarily for wheat crop. The data was divided into three sets, namely, training,
validation and testing, with a percentage of 70, 15 and 15 respectively. The values of the dataset were
facilely accountable with Arbitrary Forest Relegation and Regression. The rainfall estimation involves
many parameters highly dynamic and prone to transmutations in authentic time and ergo it made an
artificial neural network [16] and support vector machines inefficient comparatively [17]. The
conception is to design a deployable model and ergo desultory forest regression proved to be more
adaptive to dynamic genuine time data and gives proximately precise results for weather presage needed
to determine the whether it would rain or not [5]. Once it has been determined that the rainfall would
occur or not the next challenge is to presage the amplitude of rainfall which would occur. This majorly
deals with the historic precipitation data available for different regions and the model has to acclimate to
a plethora of such historic data spread over a wide duration. For this, we took the historic data available
for rainfall for different states provided by the Regime of India for research purposes. The rainfall data
involved non-linear patterns and has varied intensities associated with it. The system is designed to
acclimate to any state and withal updates the dataset with the incipient rainfall values which is being
recorded with the avail of the Raspberry Pi. The overall implementation of the system makes it highly
dynamic and more precise as well as adaptable to any region with onset of time of deployment. The
rainfall estimation in India involved monthly and yearly patterns which were studied through utilization
of Autoregressive Integrated Moving Average Time Series popularly kenned as the ARIMA Model
utilized for rainfall estimation from historic data [10]. The results were found to be as expected with
variations in intensity of rainfall and non-linear patterns visually examined uniquely for each state.
 The dataset used is described as follows:

The details of the rainfall patterns were determined through the time series [18] and the probability of the
rainfall was determined already combining both of which we get the desired output.

 Estimation of Soil Moisture Content required

The decision system takes into account the water level content required by the crop to calculate its
irrigation requisites. In traditional applications, these decisions are taken by the farmer predicated on their
experience or an expert agricultural expert. Information from sundry sources is accumulated; these include
weather statistics, crop and soil properties, moisture content data accumulated from soil sensors deployed
in the fields, and moisture influencing factors such as evapotranspiration [11]. An agricultural expert
analyses this data to decide an estimate of the quantity of water that the crop would require on a particular
day. Predicated on this conception, our autonomous irrigation system is developed. We accumulate
historical data which consists of decisions taken by an agricultural expert for the water content required
by a crop predicated on available information and current crop condition. The accumulated information is
analysed utilizing machine learning to presage the water requisites for the crop in authentic-time. Decision
history by an agricultural expert is utilized to evaluate our system’s performance. The decisions taken by
the machine learning model are evaluated against those taken by the agricultural expert. The machine
learning system is to be trained with historical data and decision reports of the agricultural expert, taking
into account the kenned water level requisite of the particular crop and sundry authentic-time factors
mentioned anteriorly. The aim of the machine learning model is to be as proximate as possible to the
decisions taken by an agronomist, which are utilized as ground truth for evaluation. We tested sundry
machine learning models to achieve the best performance with the minimum indispensable computation
required. Figure illustrates a schematic representation of the machine learning system.

The machine learning predicated decision system, when trained with precise data, can provide a precise
estimate of the irrigation requisites of the crop in consideration, given the required authentic-time
information including the crop and soil condition. Table 2 shows the set of possible factors to be
considered as input to the decision model. Evapotranspiration is derived from the words “evaporation”
and “transpiration”. It refers to the process of moisture eluding from the soil and crop to the atmosphere.
Scrutinizing the evapotranspiration rates of a crop is essential to estimating its irrigation requisites [19].

The water to be irrigated to the crop is the magnitude of water level (moisture content) required to
compensate for the evapotranspiration moisture loss from the field. Hence, ascertaining the
evapotranspiration of a crop field is a major deciding factor in estimating the irrigation requisites of the
crop. The FAO Penman-Monteith formula can be used to calculate the reference crop evapotranspiration
(ET0) on a daily basis, using information from weather stations or alternatively weather sensors [20]. The
crop evapotranspiration can be calculated under standard conditions using the single crop coefficient
formula given below:
ETc = K. ET0
Here Kc denotes the crop coefficient. It depends on various factors such as the crop type, crop
evaporation, and crop growth stage and weather information.

We’ve utilized the Partial Least Square Regression (PLSR) [21] algorithm to design the machine learning
decision model. It is a statistical algorithm that identifies fundamental cognations between input and
output variables. Predictor (input) variables, denoted by X, are defined as the visually examined variables
that are quantified for providing input to the decision model. Replication variables, denoted by Y, are the
outputs that must be estimated provided the input.

We utilize the PLSR technique among other regression algorithms since it efficiently tackles cases
when the number of inputs is much higher than the number of output variables, the outputs are
strepitous and there subsists a high probability of having multicollinearity among the input
variables. The multicollinearity quandary occurs when input variables are highly correlated, due
to redundancy between meteorological factors and soil sensors data. We ascertain that all of these
factors appear in our irrigation decision system.
 The training set {X, Y} of S samples is used to train the PLSR model. It comprises of the predictor
input matrix X=[x1, ..xi , ..xS] T and the response output matrix Y= [y1, ..yi, ..yS] T. Here xi is a
vector of N elements which contain all the soil sensor and weather data measured at a particular
day i. Since in this model, the output is only the time of irrigation recommended for that day, yi is a
scalar, containing the variable to be estimated at a given day i.

The trained PLSR model takes the predictor matrix as input which contains the soil and weather
information for a particular day, and predicts the soil moisture percentage required for the crop.
This soil moisture percentage helps us to calculate the minutes of irrigation required as a function
of the area of the crop field and the power of the electric motor being used.

PARTIAL LEAST SQUARE REGRESSION (PLSR) ALGORITHM (AI ALGORITHM

USED IN OUR PROJECT):-

The goal of pls regression is to predict Y from X and to describe their common structure. When Y is a vector and
X is full rank, this goal could be accomplished using ordinary multiple regression. When the number of predictors
is large compared to the number of observations, X is likely to be singular and the regression approach is no
longer feasible (i.e., because of multicollinearity). Several approaches have been developed to cope with this
problem. One approach is to eliminate some predictors (e.g., using stepwise methods) another one, called principal
component regression, is to perform a principal component analysis (pca) of the X matrix and then use the
principal components of X as regressors on Y. The orthogonality of the principal components eliminates the
multicolinearity problem. But, the problem of choosing an optimum subset of predictors remains. A possible
strategy is to keep only a few of the first components. But they are chosen to explain X rather than Y, and so,
nothing guarantees that the principal components, which “explain” X, are relevant for Y. By contrast, pls
regression finds components from X that are also relevant for Y. Specifically, pls regression searches for a set of
components (called latent vectors) that performs a simultaneous decomposition of X and Y with the constraint
that these components explain as much as possible of the covariance between X and Y. This step generalizes pca.
It is followed by a regression step where the decomposition of X is used to predict Y.

Simultaneous decomposition of predictors and dependent variables

Pls regression decomposes both X and Y as a product of a common set of orthogonal factors and a set of specific
loadings. So, the independent variables are decomposed as X = TPT with TTT = I with I being the identity matrix
(some variations of the technique do not require T to have unit norms). By analogy with pca T is called the score
matrix, and P the loading matrix (in pls regression the loadings are not orthogonal). Likewise, Y is estimated as
Yb = TBCT where B is a diagonal matrix with the “regression weights” as diagonal elements (see below for more
details on these weights). The columns of T are the latent vectors. When their number is equal to the rank of X,
they perform an exact decomposition of X. Note, however, that they only estimate Y. (i.e., in general Yb is not
equal to Y).

Pls regression and covariance

The latent vectors could be chosen in a lot of different ways. In fact in the previous formulation, any set of
orthogonal vectors spanning the column space of X could be used to play the rˆole of T. In order to specify T,
additional conditions are required. For pls regression this amounts to finding two sets of weights w and c in order
to create (respectively) a linear combination of the columns of X and Y such that their covariance is maximum.
Specifically, the goal is to obtain a first pair of vectors t = Xw and u = Yc with the constraints that wTw = 1, t Tt
= 1 and t To be maximal. When the first latent vector is found, it is subtracted from both X and Y and the procedure
is re-iterated until X becomes a null matrix (see the algorithm section for more).

A pls regression algorithm

The properties of pls regression can be analyzed from a sketch of the original algorithm. The first step is to create
two matrices: E = X and F = Y. These matrices are then column centered and normalized (i.e., transformed into
Zscores). The sum of squares of these matrices are denoted SSX and SSY . Before starting the iteration process,
the vector u is initialized with random 2 values. (in what follows the symbol ∝ means “to normalize the result of
the operation”).
Step 1. w ∝ ETu (estimate X weights).
Step 2. t ∝ Ew (estimate X factor scores).
Step 3. c ∝ F Tt (estimate Y weights).
Step 4. u = Fc (estimate Y scores).
If t has not converged, then go to Step 1, if t has converged, then compute the value of b which is used to predict
Y from t as b = t Tu, and compute the factor loadings for X as p = ETt. Now subtract (i.e., partial out) the effect
of t from both E and F as follows E = E−tpT and F = F−btcT. The vectors t, u, w, c, and p are then stored in the
corresponding matrices, and the scalar b is stored as a diagonal element of B. The sum of squares of X
(respectively Y) explained by the latent vector is computed as p Tp (respectively b 2 ), and the proportion of
variance explained is obtained by dividing the explained sum of squares by the corresponding total sum of squares
(i.e., SSX and SSY ). If E is a null matrix, then the whole set of latent vectors has been found, otherwise the
procedure can be reiterated from Step 1 on.

FORMULA USED

With one independent variable, we may write the regression equation as:

Where Y is an observed score on the dependent variable, a is the intercept, b is the slope, X is the observed
score on the independent variable, and e is an error or residual.

We can extend this to any number of independent variables:

 (3.1)

Note that we have k independent variables and a slope for each. We still have one error and one intercept. Again
we want to choose the estimates of a and b so as to minimize the sum of squared errors of prediction. The
prediction equation is:

(3.2)

Finding the values of b (the slopes) is tricky for k>2 independent variables, and you really need matrix algebra
to see the computations. It's simpler for k=2 IVs, which we will discuss here. But the basic ideas are the same
no matter how many independent variables you have. If you understand the meaning of the slopes with two
independent variables, you will likely be good no matter how many you have.

For the one variable case, the calculation of b and a was:

For the two variable case:

and

At this point, you should notice that all the terms from the one variable case appear in the two variable case. In
the two variable case, the other X variable also appears in the equation. For example, X2 appears in the equation
for b1. Note that terms corresponding to the variance of both X variables occur in the slopes. Also note that a
term corresponding to the covariance of X1 and X2 (sum of deviation cross-products) also appears in the
formula for the slope.

The equation for a with two independent variables is:

This equation is a straight-forward generalization of the case for one independent variable.
EQUATION USED TO CALCULATE SOIL MOISTURE REQUIRED BY THE CROP AT A GIVEN
INSTANCE:

The FAO Penman-Monteith method to estimate ETo

Tabulated Kc values
The below given Table lists typical values for Kc ini, Kc mid and Kc end for various agricultural crops. The
coefficients presented are organized by group type (i.e., small vegetables, legumes, cereals, etc.) to assist in
locating the crop in the table and to aid in comparing crops within the same group. There is usually close similarity
in the coefficients among the members of the same crop group, as the plant height, leaf area, ground coverage
and water management are normally similar.

The coefficients in Table integrate the effects of both transpiration and evaporation over time. The effects of the
integration over time represent an average wetting frequency for a 'standard' crop under typical growing
conditions in an irrigated setting. The values for Kc during the initial and crop development stages are subject to
the effects of large variations in wetting frequencies and therefore refinements to the value used for Kc ini should
always be made. For frequent wettings such as with high frequency sprinkler irrigation or rainfall, the values for
Kc ini may increase substantially.
FLOWCHART OF OUR PROJECT MODEL METHODOLOGY:
ARCHITECTURE AND DESIGN OF OUR MODEL
BACKEND CLOUD DATABASE:
HARDWARE DEVICE MODEL:-
ANDROID APPLICATION CONNECT ED TO IOT DEVICE VIA CLOUD:
SOURCE CODE OF OUR MODEL:-
#include <ArduinoJson.h>
#include <ESP8266WiFi.h>
#include <WiFiClientSecure.h>
#include <ESP8266WebServer.h>
#include <ESP8266HTTPClient.h>
#include <FirebaseArduino.h> // firebase library

#define FIREBASE_HOST "smartgardener-833c2.firebaseio.com" // the project

name address from firebase id
#define FIREBASE_AUTH "f1BiVsro8Ft3oDXKfu5XGXyssYNVd8Q2BauvthRy" // the
secret key generated from firebase

//SENSOR AREA
#include "DHT.h" // including the library of DHT11 temperature and humidity sensor
#define DHTTYPE DHT11 // DHT 11

#define dht_dpin 0
DHT dht(dht_dpin, DHTTYPE);
//Variable
float sensorMoisture;
float h;
float t;
float moisture_percentage;
int sensor_analog;
float ET0;
float KCET0;
float Kc;
 float PrecipProb;
float G;
float SumY;
float tempMin;
float tempMax;
float Pres;
float rh;
float solarRad;
float windSpd;
float dni;
/* Set these to your wifi credentials. */
const char *ssid = "Bsharma"; //ENTER YOUR WIFI SETTINGS
const char *password = "9993936448";

const char *host = "irrisat-cloud.appspot.com";

const int httpsPort = 443; //HTTPS= 443 and HTTP = 80

//SHA1 finger print of certificate use web browser to view and copy
const char fingerprint[] PROGMEM = "e0 5c 17 d7 4a 85 68 8d d4 69 ba d3 d5 da 5c 89 cf 2d b0
29";
//============================================
// Power on setup
//============================================

void setup() {
delay(1000);
digitalWrite(4,HIGH);
 pinMode(4, OUTPUT);
Serial.begin(115200);
WiFi.mode(WIFI_OFF); //Prevents reconnection issue (taking too long to connect)
delay(1000);
WiFi.mode(WIFI_STA); //Only Station No AP, This line hides the viewing of ESP as wifi
hotspot
WiFi.begin(ssid, password); //Connect to your WiFi router
Serial.println("");

Serial.print("Connecting");
// Wait for connection
while (WiFi.status() != WL_CONNECTED) {
delay(500);
Serial.print(".");
}
//If connection successful show IP address in serial monitor
Serial.println("");
Serial.print("Connected to ");
Serial.println(ssid);
Serial.print("IP address: ");
Serial.println(WiFi.localIP()); //IP address assigned to your ESP
Firebase.begin(FIREBASE_HOST, FIREBASE_AUTH);
}
//=================================================================
// Main Program Loop
//=================================================================
void loop() {
//================IRRISTAT API============================//
HTTPClient httpAgro; //Declare an object of class HTTPClient

httpAgro.begin("http://api.weatherbit.io/v2.0/forecast/agweather?lat=23.2756&lon=77.4560&k
ey=10fe4d3c09b64cb18e44c87dd9003578");//("http://api.weatherbit.io/v2.0/current?lat=23.275
6&lon=77.4560&key=10fe4d3c09b64cb18e44c87dd9003578"); //Specify request destination
int httpAPICode = httpAgro.GET(); //Send the request
if (httpAPICode > 0) { //Check the returning code
String payload = httpAgro.getString(); //Get the request response payload
int ind = payload.indexOf('}');
///////////////////JSON PARSINNG///////////////////////////////
Serial.println("\n");
Serial.println(payload.substring(0,ind+1)); //Print the response payload
String final =payload.substring(0,ind+1)+"]}";
DynamicJsonBuffer jsonBuffer(907);
char json[final.length() + 1];
final.toCharArray(json, final.length() + 1);
JsonObject& root = jsonBuffer.parseObject(json);
// Test if parsing succeeds.
if (!root.success()) {
Serial.println("parseObject() failed");
return;
}
ET0 = root["data"][0]["evapotranspiration"];
Kc = Firebase.getFloat("/Kc");
KCET0 = ET0*Kc;
Firebase.setFloat("/KCET0",KCET0);
 PrecipProb = root["data"][0]["precip"];
G = root["data"][0]["soilt_0_10cm"];
Firebase.pushFloat("/ET0",ET0);
Firebase.pushFloat("/GSoilFlux",G);
Firebase.setFloat("/CurrentET0",ET0);
SumY = Firebase.getFloat("/SumY");
SumY = SumY +ET0;
Firebase.setFloat("/SumY",SumY);
Firebase.setFloat("/PrecipProbability",PrecipProb);
}
httpAgro.end(); //Close connection
//=============OpenWeatherAPI====================//
HTTPClient http; //Declare an object of class HTTPClient

http.begin("http://api.openweathermap.org/data/2.5/weather?lat=23.2756&lon=77.4560&appid
=8f9bc27b855db221dae3a2aa4880ce73");//Specify request destination
int httpCode = http.GET(); //Send the request

if (httpCode > 0) { //Check the returning code

String openW = http.getString(); //Get the request response payload

Serial.println(openW);

///////////////////JSON PARSINNG///////////////////////////////
DynamicJsonBuffer jsonBufferO(782);
char jsonO[openW.length() + 1];
openW.toCharArray(jsonO, openW.length() + 1);

JsonObject& rootO = jsonBufferO.parseObject(jsonO);

// Test if parsing succeeds.

if (!rootO.success()) {
Serial.println("parseObject() failed");
return;
}
const char* desc = rootO["weather"][0]["description"];
Firebase.setString("/CurrentWeather",desc);

tempMin = (rootO["main"]["temp_min"]);
tempMax = (rootO["main"]["temp_max"]);
tempMin = tempMin - 273;
tempMax = tempMax - 273;

Firebase.pushFloat("/dataTempMin",tempMin);
Firebase.pushFloat("/dataTempMax",tempMax);
Firebase.setFloat("/DisplayTempMin",tempMin);
Firebase.setFloat("/DisplayTempMax",tempMax);

float disPres = rootO["main"]["pressure"];

float disHum = rootO["main"]["humidity"];
Firebase.setFloat("/DisplayPressure",disPres);
Firebase.setFloat("/DisplayHumidity",disHum);
}
http.end(); //Close connection
//=========OPENWEATHERCLOSE_CONNECTION====================//

//===============WEATHERBIT API==================//

HTTPClient httpWeatherBit; //Declare an object of class HTTPClient

httpWeatherBit.begin("http://api.weatherbit.io/v2.0/current?lat=23.2756&lon=77.4560&key=10
fe4d3c09b64cb18e44c87dd9003578"); //Specify request destination
int httpBitCode = httpWeatherBit.GET(); //Send the
request
if (httpBitCode > 0) { //Check the returning code
String WeatherBit = httpWeatherBit.getString(); //Get the request response payload
Serial.println(WeatherBit);
///////////////////JSON PARSINNG///////////////////////////////
DynamicJsonBuffer jsonBufferW(782);

char jsonW[WeatherBit.length() + 1];

WeatherBit.toCharArray(jsonW, WeatherBit.length() + 1);
JsonObject& rootW = jsonBufferW.parseObject(jsonW);
// Test if parsing succeeds.
if (!rootW.success()) {
Serial.println("parseObject() failed");
return;
}
Pres = rootW["data"][0]["pres"];
rh = rootW["data"][0]["rh"];
 solarRad = rootW["data"][0]["solar_rad"];
windSpd = rootW["data"][0]["wind_spd"];
dni = rootW["data"][0]["dni"];
Firebase.pushFloat("/PressureCalc",Pres);
Firebase.pushFloat("/RelativeHumidity",rh);
Firebase.pushFloat("/SolarRad",solarRad);
Firebase.pushFloat("/WindSpeed",windSpd);
Firebase.setFloat("/DisplayWindSpeed",windSpd);
Firebase.pushFloat("/DirectNSI",dni);
}

httpWeatherBit.end(); //Close connection

//===================PLSR ALGO===================================//
float tempMean = (tempMin+tempMax)/2;
float p = (17.27*tempMean)/(tempMean+273.3);
float del = (4098*( 0.6108 * pow(2.71828,p)))/pow((tempMean + 237.3),2);
float gama = 0.000665*Pres;
float Rn = solarRad;
float u2 = windSpd;
float p1 = (17.27*tempMax)/(tempMax+273.3);
float p2 = (17.27*tempMin)/(tempMin+273.3);
float eTmax = 0.6108*(pow(2.71828,p1));
float eTmin = 0.6108*(pow(2.71828,p2));
float es = (eTmax + eTmin)/2;
float ea = (eTmin)*(rh/100);
float X1 = (del*(Rn-G))/(del + (gama*(1+ 0.34*u2)));
float X2 = (gama*u2*(es-ea))/((tempMean+273)*(del + (gama*(1+ 0.34*u2))));
Firebase.pushFloat("/X1",X1);
Firebase.pushFloat("/X2",X2);
float Y = ET0;
float X1Y = X1*Y;
float X2square = X2*X2;
float X1X2 = X1*X2;
float X2Y = X2*Y;
float X1square = X1*X1;
float SumX1 = Firebase.getFloat("/SumX1");
SumX1 = SumX1+X1;
float SumX2 = Firebase.getFloat("/SumX2");
SumX2 = SumX2+X2;
float SumX1Y = Firebase.getFloat("/SumX1Y");
SumX1Y=SumX1Y+X1Y;
float SumX2square = Firebase.getFloat("/SumX2square");
SumX2square=SumX2square+X2square;
float SumX1X2 = Firebase.getFloat("/SumX1X2");
SumX1X2=SumX1X2+X1X2;
float SumX2Y = Firebase.getFloat("/SumX2Y");
SumX2Y=SumX2Y+X2Y;
float SumX1square = Firebase.getFloat("/SumX1square");
SumX1square=SumX1square+X1square;

Firebase.setFloat("/SumX1",SumX1);
Firebase.setFloat("/SumX2",SumX2);
Firebase.setFloat("/SumX1Y",SumX1Y);
Firebase.setFloat("/SumX2square",SumX2square);
Firebase.setFloat("/SumX1X2",SumX1X2);
Firebase.setFloat("/SumX2Y",SumX2Y);
Firebase.setFloat("/SumX1square",SumX1square);

int c = Firebase.getInt("/count");
float b1 = ((SumX2square*SumX1Y)-(SumX1X2*SumX2Y))/((SumX1square*SumX2square)-
pow(SumX1X2,2));
float b2 = ((SumX1square*SumX2Y)-(SumX1X2*SumX1Y))/((SumX1square*SumX2square)-
pow(SumX1X2,2));
float b0 = (SumY/c) - (b1*(SumX1/c)) - (b2*(SumX2/c));

float YFinal = b0 +(b1*X1)+(b2*X2);

float ET0F = (((0.408*X1)+(900*X2))*2)/10;

Firebase.pushFloat("/ET0F",ET0F);
Firebase.pushFloat("/b1",b1);
Firebase.pushFloat("/b2",b2);
Firebase.pushFloat("/b0",b0);
sensor_analog = analogRead(A0);
moisture_percentage = ( 100 - ( ((sensor_analog-300)/724.00) * 100 ) );
Serial.print("Moisture Percentage = ");
Serial.print(moisture_percentage);
Serial.print("%\n\n");

h = dht.readHumidity();
t = dht.readTemperature();
sensorMoisture= ((float)(moisture_percentage)/10);
delay(5000);
if (sensorMoisture<6 && sensorMoisture<KCET0 && PrecipProb<KCET0){
Serial.println("\MotorON");
digitalWrite(4,LOW);
delay(2000);
}
while(sensorMoisture<7 && sensorMoisture<KCET0 && PrecipProb<KCET0){
sensor_analog = analogRead(A0);
moisture_percentage = ( 100 - ( ((sensor_analog-300)/724.00) * 100 ) );
sensorMoisture= ((float)(moisture_percentage)/10);
Serial.println("Updated Moisture:");
Serial.println(sensorMoisture);
delay(5000);
}
digitalWrite(4,HIGH);
Serial.println("\MOTOR OFF");
Firebase.setInt("/count",c+1);
delay(1800000);
}

COMPARISON BETWEEN TWO BEST ALGORITHMS FOR REQUIRED SOIL

MOISTURE DETECTION
1) Artificial neural network model (ANN)
An ANN consists of a large number of highly interconnected processing elements (nodes or units), which
simulate basic functions of biological neurons. Fig. 4 illustrates a simple oneneuron model within an
artificial network, where an input vector is passed through the neuron for providing an output value. A
multiple-layer system consists of input, hidden and output layers. The theory and mathematical basis of
ANNs have been extensively described in Haykin (1999) and Bishop (1995). In this study, a multiple-
layer feed-forward backpropagation network with three layers was used: input, hidden and output layer.
This type of network generally provides better performances in comparison to other types (Hornik et al.,
1989). Tan– sigmoid transfer functions (non-linear) and linear transfer functions were selected for the
 hidden and output layers, respectively. Using a tan– sigmoid function in the hidden layer allows to
approximate only non-linear relations present between input and output layers (Haykin, 1999). The
number of neurons in the hidden layer can be defined using a formula recommended by Fletcher and Goss
(1993) or using a trial- and-error approach as suggested by Chang et al. (2004). The number of neurons in
the hidden layer is of great importance, as too many neurons may cause over-fitting problems (Huang &
Foo, 2002). For improving network generalization, Demuth and Beale (2004) recommended to use a
network that is just large enough to provide an adequate fit between predictor and response variables. To
avoid over-fitting problems and to provide an effective means to stop the time demanding training phase,
the so called “early stopping” approach has been used (Demuth & Beale, 2004). The network input layer
used in this study relates to reflectance, while the network output layer relates to soil salinity. Data
preprocessing techniques (centering or standardizing to a mean of zero and standard deviation of one) and
principal component analysis (PCA) were applied to the inputs to normalize the reflectance data and to
reduce the data dimension. The numbers of neurons used for training the networks were varied
systematically between five and 13 to allow subsequent selection of the most appropriate network size
based on the performance on the test data set (Adeloye & De Munare, 2005). For ANN modeling, the
computer software MatLab and the Neural Network Toolbox were used (Demuth & Beale, 2004). In each
scale of the study, each data set was divided into three subsets, one for training (half of input data), one
for validation (a quarter) and one for testing (a quarter of input data). The Levenberg–Marquardt algorithm
(Demuth & Beale, 2004), which provides a fast optimization, was used for network training. The
performance of a trained network was assessed by comparing the mean squared error (MSE) and root
mean squared error (RMSE) calculated from training, validation and testing data subsets. Only a training
data subset is used for updating the network weights and biases. During training, error with respect to
validation data subset is monitored. When the validation error increases for a specified number of
iterations, the training is stopped. Error with respect to testing data subset is not monitored during training,
but is quantified to assess the final performance of a trained ANN model.
2) Quantification of model performance and estimation errors between PLSR and ANN
Algorithms
To quantify performance of salinity–reflectance models based on PLSR and ANN methods, various parameters
between estimated values (Y′) and independent reference measurements (Y) were calculated (Table 2): root mean
square error (RMSE), relative RMSE (RMSE%), mean and relative mean absolute errors (MAE and MAE%). Both
MAE and RMSE indicate absolute estimation errors, but RMSE is more sensitive to outliers and thus, RMSE% and
MAE% are also calculated. To analyse the goodness of estimated values versus the observed values, three other
model performance parameters were calculated (Miehle et al., 2006; Nash & Sutcliffe, 1970; Williams, 2001):
coefficient of determination (R2), model efficiency (ME) and ratio of prediction to deviation (RPD). Detailed
descriptions and definitions of these model performance parameters are given in Taylor (1997) and Cacuci (2003).
The equations for quantification of the model performance parameters are given in Table 2. The R2 and RPD
indicate strength of statistical correlation between measured values and predicted values both for PLSR and ANN
predictive models. A predictive model is accurate if R2 and RPD values are higher than 0.91 and 2.5, respectively
(Williams, 2001). R2 between 0.82–0.9 and RPD higher than 2 indicate a good prediction, whereas the R2 of an
approximate prediction is considered to lie between 0.66–0.81 with a RPD of higher than 1.5. R2 between 0.5 and
0.65 indicates a poor relationship. MAE and ME indices were also used to examine the performance of the PLSR
and ANN predictive models. MAE and ME values indicate degree of agreement between measured and predicted
values; hence, they provide a measure of prediction model efficiency (Nash & Sutcliffe, 1970). Larger values MAE%
and RMSE%, thus mean high prediction errors while ME values close to 1 indicate small errors in prediction values.
To examine the performance of PLSR-calibration models RMSE and RPD values is analysed. Generalization of the
 ANN-training, validation and test models was analysed using the R2 and RMSE values calculated from each data
sets (training, validation and test).

RESULTS AND DISCUSSION

The system designed is an astute irrigation solution predicated on artificial astuteness which is makes utilization
of the soil moisture content and the moisture requisites of the crop to make the entire process of irrigation
automatic. Its core benefit is its efficiency and economic feasibility. The main conception abaft the rainfall
estimation proposed is to get the virtually correct estimation of the rainfall of a concrete local region as well as
the annual rainfall to get data for future estimation of the rainfall in states. The variation in the non-linear
patterns of rainfall for India over the historic data is scrutinized and patterns visually examined in the historic
period from 1871 to 2017 are as follows:

The annual rainfall presaged for the state of Punjab utilizing Time Series ARIMA Model is 545 mm for the year
2015 which is proximately to the pristine value of 549.5 mm with a precision of 83%. This is further verified
with the monthly rainfall presage in the state. The system prognosticates the probability of rain through this and
thus if the probability of rain emerges to be higher than 0.5 then the motor is not switched. If the probability of
rain is lesser than 0.5 then the soil moisture requisite is calculated provided the soil moisture sensor data thereby
switching on the motor for only the required duration automatically.

CONCLUSION
The autonomous irrigation system we developed uses Artificial Intelligence learning and predictive algorithms
to integrate perspicacity to subsisting concept of automatic irrigation systems. The methods discussed in this
paper can avail increase irrigation efficiency while decrementing effort required and avails water conservation
compared to current irrigation methods. The system currently depends on weather station information for its
calculation. This dependency can be superseded with on-premise sensors for deployment in rural areas widely
found in the Indian subcontinent and arid regions where water is available in constrained quantity.

REFERENCES
[1] Edordu C. and Sacks L., “Self Organizing Wireless Sensor Networks as a Land Management Tool in
Developing Countries: A Preliminary Survey,” In Proceedings of the 2006 London Communications
Symposium, September 2006, Communications Engineering Doctorate Centre, London, UK.

[2] Kim Y., Evans R.G. and Iversen W.M., “Remote Sensing and Control of an Irrigation System Using a
Distributed Wireless Sensor Network,” Instrumentation and Measurement, IEEE Transactions on, vol.57, no.7,
pp.1379-1387, July 2008. http://dx.doi.org/10.1109/TIM.2008.917198

[3] Joaquín G, Juan F , Alejandra N.G, and Miguel Ángel, “Automated Irrigation System Using a Wireless
Sensor Network and GPRS Module”, IEEE Transactions On Instrumentation and Measurement, Vol.63, no.1,
pp.166-176, 2013
[4] Karandeep K, “Machine Learning : Applications in Indian Agriculture”, International Journal of Advanced
Research in Computer and Communication Engineering, Vol.5, no.4, pp.342- 344, 2016.
[5] Washington O and Joseph O, ”Machine Learning Classification Technique for Famine Prediction”.
Proceedings of the World Congress on Engineering 2011 Vol II WCE 2011, July 6 - 8, London, U.K, 2011.

[6] Snehal S, Sandeep V.R, ”Agricultural Crop Yield Prediction Using Artificial Neural NetworkApproach”.
International Journal of Innovative Research in Electrical, Electronic, Instrumentation and Control Engineering,
Vol. 2, Issue 1, January 2014.

[7] Kumar R, Singh M .P, Prabhat K, and Singh J.P. "Crop Selection Method to Maximize Crop Yield Rate
Using Machine Learning Technique." Proceedings of International Conference on Smart Technologies and
Management for Computing, Communication, Controls, Energy and Materials (ICSTM) ,2015.

[8] Rumpf, T., A.-K. Mahlein, U. Steiner, E.-C. Oerke, H.-W. Dehne, and L. Plümer. "Early Detection and
Classification of Plant Diseases with Support Vector Machines Based on Hyperspectral Reflectance."
Computers and Electronics in Agriculture, Vol. 74, no.1, pp.91-99, 2010.

[9] Carles Antón-Haro, Thierry Lestable, Yonghua Lin, Navid Nikaein, Thomas Watteyne, Jesus AlonsoZarate,
“Machine-to-machine: An emerging communication paradigm. Trans. on Emerging Telecommunications
Technologies”, Volume 24, Issue 4, pages 353–354, June 2013. DOI: http://dx.doi.org/10.1002/ett.2668

[10] Koushik Anand, Jayakumar C, Mohana Muthu and Sridhar A, “Automatic Drip Irrigation using Fuzzy
Logic and Mobile Technology”, Proceedings of Technological Innovation in ICT for Agriculture and Rural
Development, 2015.

[11] Kait L.K., Kai C.Z., Khoshdelniat R., Lim S.M., and Tat E.H., “Paddy Growth Monitoring with Wireless
Sensor Networks,” Proceedings of Intelligent and Advanced Systems, Kuala Lumpur, Malaysia, pp.966- 970,
2007. http://dx.doi.org/10.1109/ICIAS.2007.4658529
[12] Kim Y., Evans R.G. and Iversen W.M., “Remote Sensing and Control of an Irrigation System Using a
Distributed Wireless Sensor Network,” Instrumentation and Measurement, IEEE Transactions on, vol.57, no.7,
pp.1379-1387, July 2008. http://dx.doi.org/10.1109/TIM.2008.917198

[13] Wall R.W. and King B.A., “Incorporating plug and play technology into measurement and control systems
for irrigation management”, presented at the ASAE/CSAE Annu. Int. Meeting, Ottawa, ON, Canada, Aug. 2004

[14] Yang W., Liusheng H., Junmin W. and Hongli X., “Wireless Sensor Networks for Intensive Irrigated
Agriculture,” Consumer Communications and Networking Conference, 2007. CCNC 2007. 4th IEEE, pp.197-
201, Las Vegas, Nevada, Jan. 2007. http://dx.doi.org/10.1109/CCNC.2007.46

[15] Konstantinos K., Apostolos X., Panagiotis K. and George S., “Topology Optimization in Wireless Sensor
Networks for Precision Agriculture Applications,” Sensor Technologies and Applications, 2007. SensorComm
2007. International Conference on, pp.526-530, Valencia, Spain, 14-20 Oct. 2007.
http://dx.doi.org/10.1109/SENSORCOMM.2007.101

[16] Masuki, K. F. Ga, Kamugisha, Rb, Mowo, J. Gc, Tanui, Jc, Tukahirwa, Ja. Mogoi, Jc. and Adera E.O,
“Role of mobile phones in improving communication and information delivery for agricultural development”,
ICT and Development - Research Voices from Africa. International Federation for Information Processing
(IFIP), Technical Commission 9 – Relationship Between Computers and Society. Workshop at Makerere
University, Uganda. 22-23 March 2010
[17] Suyash S P and Sandeep A T, “Early Detection of Grapes Disease using Machine Learning and IoT”,
Proceedings of Second International Conference on Cognitive Computing and Information Processing, Mysore,
India, 2016

[18] Yue L, Long M and Ooi S K, “Prediction of Soil Moisture based on Extreme Learning Machine for an
Apple Orchard”, Proceedings of 3rd IEEE International Conference on Cloud Computing and Intelligent
System”, Shenzhen, Hong Kong, 2014.

[19] Biswas S, Saunshi A, Sarangi S and Pappula S, “Random Forest based Classification of Diseases in Grapes
from Images Captured in Uncontrolled Envrionemnt”, Proceedings of IET International Conference on Signal
Processing, 2016.

[20] Rakesh K, Singh M.P, Prabhat K and Singh J.P, “Crop Selection Method to maximize Crop Yield Rate
using Machine Learning Technique”, Proceedings of International Conference on Smart Technologies and
Management for Computing, Communication, Controls, Energy and Materials, Chennai, India, 2015.
 SMART IRRIGATION USING ARTIFICIAL INTELLIGENCE AND IOT
18BCE0107, 18BCE0154,18BCE0172|PROF. BHULAKSHMI BONTHU| SCOPE

Introduction Results
we provide an Automatic Irrigation System predicated on Artificial Perspicacity The system designed is an astute irrigation solution predicated on artificial
and Internet of Things, which can autonomously irrigate fields utilizing soil astuteness which makes utilization of the soil moisture content and the moisture
moisture data. The system is predicated on prognostication algorithms which requisites of the crop to make the entire process of irrigation automatic. Its core
make utilization of historic weather data to identify and presage rainfall patterns benefit is its efficiency and economic feasibility. The main conception abaft the
and climate changes. rainfall estimation proposed is to get the virtually correct estimation of the
rainfall of a concrete local region as well as the annual rainfall to get data for
Motivation future estimation of the rainfall in states. The variation in the non-linear patterns
of rainfall for India over the historic data is scrutinized and patterns visually
Farmers primarily depend on personal monitoring and their experience in examined in the historic period from 1871 to 2017 are as follows:
irrigating the fields, and as a result, irrigation becomes largely inefficient and
eccentric, therefore our device will help them in irrigation.

SCOPE of the Project

Agriculture is the primary source of livelihood for about 58 percent of India’s
population. 70 % of rural India is dependent on agriculture for money. There is
clearly a huge market for irrigational products.

To achieve this, we make utilization of Node MCUs, and soil moisture sensors,
placed inside waterproof boxes and spread evenly throughout the area to be
irrigated. All these nodes are connected to cloud via Wi-Fi connection

Methodology
The system designed consists briefly of two major components:

• Artificial Intelligence to predict the amount of rainfall and the soil moisture
content.

• Implementation of system using Internet of Things.

Distributed Nodes

Each node is comprised of an ESP8266 [8] predicated microcontroller board

(NodeMCU) and an array of analog sensors. These sensors are acclimated to fetch
information about weather conditions, which is converted to digital 10-bit
integers utilizing the Analog to Digital converters present on each NodeMCU.
The nodes are connected to the central controller
through a shared Wi-Fi access point, over the IPv4
protocol. A client-server architecture is utilized for
communication between the nodes and the
central controller. The nodes do not communicate
with each other. This implementation utilizes a
Cloud server on the central controller. The server Farmers will be able to do the irrigation easily by
elongates a URL in the form of a web hook. The following the below steps:
nodes connect to the server via this web hook to
send data. The HTTP GET request is utilized to 1. They just have to select the type of crop they are
transmit the sensor data after post-calculation to growing in their farm and the current growth
the server. The default replication of the server stage. That's it rest our IOT device will handle.
conveys two messages to the node that sent the
request: 1. if the pump annexed to the particular RESULT ANALYSIS:
node has to be switched on or not, and 2. the Our IOT device and app are connected to a single
duration for which the pump has to be switched cloud server and can communicate through the same.
on. The duration is calculated utilizing the output
of the regression algorithm. Each node sends a IOT device finds the current environment data and
request to the server at an interval of 30 minutes. send it to cloud and also calculate the required
amount of soil moisture on the basis of current factors
PARTIAL LEAST SQUARE REGRESSION
through PLSR algorithm and at the same time updates
(PLSR) ALGORITHM (AI ALGORITHM all the information on our Android App.
USED IN OUR PROJECT):- This will conserve water and will increase the crop
The properties of pls regression can be
analyzed from a sketch of the original
production of farmers and it is easy to use device.
algorithm. The first step is to create two
matrices: E = X and F = Y. These matrices Conclusion
are then column centered and normalized
(i.e., transformed into Zscores). The sum The autonomous irrigation system we developed uses Artificial Intelligence learning
of squares of these matrices are denoted and predictive algorithms to integrate perspicacity to subsisting concept of
SSX and SSY . Before starting the iteration automatic irrigation systems. The methods discussed in this paper can avail increase
. process, the vector u is initialized with irrigation efficiency while decrementing effort required and avails water
random 2 values. (in what follows the conservation compared to current irrigation methods. The system currently depends
symbol ∝ means “to normalize the result on weather station information for its calculation. This dependency can be
of the operation”).
Step 1. w ∝ ETu (estimate X weights).
superseded with on-premise sensors for deployment in rural areas widely found in
Step 2. t ∝ Ew (estimate X factor scores). the Indian subcontinent and arid regions where water is available in constrained
Step 3. c ∝ F Tt (estimate Y weights). quantity.
Step 4. u = Fc (estimate Y scores).
References
[1] Kim Y., Evans R.G. and Iversen W.M., “Remote Sensing and Control of an Irrigation System Using a
Distributed Wireless Sensor Network,” Instrumentation and Measurement, IEEE Transactions on, vol.57,
no.7, pp.1379-1387, July 2008. http://dx.doi.org/10.1109/TIM.2008.917198
[2] Joaquín G, Juan F , Alejandra N.G, and Miguel Ángel, “Automated Irrigation System Using a Wireless
Sensor Network and GPRS Module”, IEEE Transactions On Instrumentation and Measurement, Vol.63, no.1,
pp.166-176, 2013
[3] Karandeep K, “Machine Learning : Applications in Indian Agriculture”, International Journal of Advanced
Research in Computer and Communication Engineering, Vol.5, no.4, pp.342- 344, 2016.
MICROPROCESSOR AND INTERFACING – CSE2006
J - COMPONENT

SMART IRRIGATION USING

ARTIFICIAL INTELLIGENCE AND IOT
 ABSTRACT :
● Agriculture plays a consequential role in the economy and its contribution is predicated on quantifiable crop
yield which is highly dependent upon irrigation.

● Through our project the farm can be modernized with electronic technology that continuously monitors the
conditions of plants and soil, so that the plants could be provided watering as required.

● The water pump is automatically controlled based on the values of the various environmental factors like
temperature, humidity, soil moisture and light intensity which we can measure through sensors like DHT-11
Temperature and Humidity sensor, Moisture sensor .

● Through our research work we will try to make the model intelligent by storing the previous scanned values
in database and doing the pre-perception on the basis of historic values stored in learning phase of the
working model. We are using PLSR algorithm to train our model.

OBJECTIVE
 we have developed a perspicacious system which can study the patterns of rainfall in a region, and soothsay weather
conditions in order to habituate to the geography, thus prognosticate the quantity of water needed for irrigation, to
minimize wastage and increment the crop yield
ALGORITHM USED TO SOLVE PROBLEM STATEMENT

PARTIAL LEAST SQUARE REGRESSION (PLSR) ALGORITHM (AI ALGORITHM USED IN OUR PROJECT):-

● The goal of pls regression is to predict Y from X and to describe their common structure. When Y is a vector and X is
full rank, this goal could be accomplished using ordinary multiple regression. When the number of predictors is large
compared to the number of observations, X is likely to be singular and the regression approach is no longer feasible
(i.e., because of multicollinearity).

● The trained PLSR model takes the predictor matrix as input which contains the soil and weather information for a
particular day, and predicts the soil moisture percentage required for the crop. This soil moisture percentage helps
us to calculate the minutes of irrigation required as a function of the area of the crop field and the power of the
electric motor being used.

THE EQUATIONS TO CALCULATE THE SOIL MOISTURE RESULT INTO THE

BELOW GIVEN MATHEMATICAL FORM:
ARCHITECTURE ( DESIGN, BLOCK DIAGRAM)
IOT MODULES USED
NodeMCU DHT11 Sensor Soil Moisture Sensor

1- channel Relay Module Motor Pump (mini Breadboard

DC 12V)
 GROUP MEMBER’S CONTRIBUTION

1. AYUSH SHARMA (18BCE0172)

• IOT Device build up, Arduino coding and its integration with android app
• Coded PLSR algorithm in IOT device
• Connected IOT device to cloud using Arduino IDE coding and connected it to
Android app too.

2. RITIK GUPTA ( 18BCE0154)

• Coded backend API to communicate with Historic data

• Coded API to fetch important factors for moisture content calculation.
• Worked on PLSR Algorithm in integrating it with backend API

3. NAMAN AGRAWAL (18BCE0107)

• Built Android App and connected it with IOT device

• Coded backend for cloud server (Server side code)
• Connected app and cloud together with IOT device. (Coded in Android
Studio)
CONCLUSION
● The autonomous irrigation system we developed uses Artificial Intelligence learning and predictive
algorithms to integrate perspicacity to subsisting concept of automatic irrigation systems. The methods
discussed in this ppt can avail increase irrigation efficiency while decrementing effort required and avails
water conservation compared to current irrigation methods.

● The system currently depends on sensors and weather station information for its calculation. This
dependency can be superseded with on-premise sensors for deployment in rural areas widely found in the
Indian subcontinent and arid regions where water is available in constrained quantity.
MAIN SOURCE CODE
void setup() {
delay(1000);
digitalWrite(4,HIGH);

pinMode(4, OUTPUT);

Serial.begin(115200);
WiFi.mode(WIFI_OFF); //Prevents reconnection issue (taking too long to connect)
delay(1000);
WiFi.mode(WIFI_STA); //Only Station No AP, This line hides the viewing of ESP as wifi hotspot

WiFi.begin(ssid, password); //Connect to your WiFi router

Serial.println("");

Serial.print("Connecting");
// Wait for connection
while (WiFi.status() != WL_CONNECTED) {
delay(500);
Serial.print(".");
}
//If connection successful show IP address in serial monitor
Serial.println("");
Serial.print("Connected to ");
Serial.println(ssid);
Serial.print("IP address: ");
Serial.println(WiFi.localIP()); //IP address assigned to your ESP
Firebase.begin(FIREBASE_HOST, FIREBASE_AUTH);
}
// Main Program Loop
//=======================================================================
void loop() {

//=========================================IRRISTAT API============================//
HTTPClient httpAgro; //Declare an object of class HTTPClient

httpAgro.begin("http://api.weatherbit.io/v2.0/forecast/agweather?lat=23.2756&lon=77.4560&key=10fe4d3c09b64cb18e44c87dd900
3578");//("http://api.weatherbit.io/v2.0/current?lat=23.2756&lon=77.4560&key=10fe4d3c09b64cb18e44c87dd9003578"); //Specify
request destination
int httpAPICode = httpAgro.GET(); //Send the request

if (httpAPICode > 0) { //Check the returning code

String payload = httpAgro.getString(); //Get the request response payload

int ind = payload.indexOf('}');

///////////////////JSON PARSINNG///////////////////////////////
Serial.println("\n");
Serial.println(payload.substring(0,ind+1)); //Print the response payload
String final =payload.substring(0,ind+1)+"]}";

DynamicJsonBuffer jsonBuffer(907);

char json[final.length() + 1];

final.toCharArray(json, final.length() + 1);

JsonObject& root = jsonBuffer.parseObject(json);

);
// Test if parsing succeeds. //====================PLSR ALGO====================//
if (!root.success()) {
Serial.println("parseObject() failed"); float tempMean = (tempMin+tempMax)/2;
float p = (17.27*tempMean)/(tempMean+273.3);
return; float del = (4098*( 0.6108 * pow(2.71828,p)))/pow((tempMean +
} 237.3),2);
ET0 = root["data"][0]["evapotranspiration"]; float gama = 0.000665*Pres;
Kc = Firebase.getFloat("/Kc"); float Rn = solarRad;
KCET0 = ET0*Kc; float u2 = windSpd;
Firebase.setFloat("/KCET0",KCET0); float p1 = (17.27*tempMax)/(tempMax+273.3);
PrecipProb = root["data"][0]["precip"]; float p2 = (17.27*tempMin)/(tempMin+273.3);
float eTmax = 0.6108*(pow(2.71828,p1));
G = root["data"][0]["soilt_0_10cm"]; float eTmin = 0.6108*(pow(2.71828,p2));
Firebase.pushFloat("/ET0",ET0); float es = (eTmax + eTmin)/2;
Firebase.pushFloat("/GSoilFlux",G); float ea = (eTmin)*(rh/100);
Firebase.setFloat("/CurrentET0",ET0);
SumY = Firebase.getFloat("/SumY"); float X1 = (del*(Rn-G))/(del + (gama*(1+ 0.34*u2)));
SumY = SumY +ET0; float X2 = (gama*u2*(es-ea))/((tempMean+273)*(del + (gama*(1+
Firebase.setFloat("/SumY",SumY); 0.34*u2))));
Firebase.pushFloat("/X1",X1);
Firebase.setFloat("/PrecipProbability",PrecipProb); Firebase.pushFloat("/X2",X2);
// Print values. float Y = ET0;
//Serial.println(ET0);
float X1Y = X1*Y;
} float X2square = X2*X2;
float X1X2 = X1*X2;
httpAgro.end(); //Close connection float X2Y = X2*Y;
float X1square = X1*X1;
//====================PLSR ALGO==============// float b2 = ((SumX1square*SumX2Y)-
float SumX1 = Firebase.getFloat("/SumX1"); (SumX1X2*SumX1Y))/((SumX1square*SumX2square)-
SumX1 = SumX1+X1; pow(SumX1X2,2));
float SumX2 = Firebase.getFloat("/SumX2"); float b0 = (SumY/c) - (b1*(SumX1/c)) - (b2*(SumX2/c));
SumX2 = SumX2+X2; float YFinal = b0 +(b1*X1)+(b2*X2);
float SumX1Y = Firebase.getFloat("/SumX1Y"); float ET0F = (((0.408*X1)+(900*X2))*2)/10;
SumX1Y=SumX1Y+X1Y;
float SumX2square = Firebase.getFloat("/SumX2square"); Firebase.pushFloat("/ET0F",ET0F);
SumX2square=SumX2square+X2square; Firebase.pushFloat("/b1",b1);
float SumX1X2 = Firebase.getFloat("/SumX1X2"); Firebase.pushFloat("/b2",b2);
SumX1X2=SumX1X2+X1X2; Firebase.pushFloat("/b0",b0);
float SumX2Y = Firebase.getFloat("/SumX2Y");
SumX2Y=SumX2Y+X2Y; //==========SENSOR VALUES TO CLOUD================//
float SumX1square = Firebase.getFloat("/SumX1square");
SumX1square=SumX1square+X1square; sensor_analog = analogRead(A0);
moisture_percentage = ( 100 - ( ((sensor_analog-300)/724.00) * 100
Firebase.setFloat("/SumX1",SumX1); ) );
Firebase.setFloat("/SumX2",SumX2); Serial.print("Moisture Percentage = ");
Firebase.setFloat("/SumX1Y",SumX1Y); Serial.print(moisture_percentage);
Firebase.setFloat("/SumX2square",SumX2square); Serial.print("%\n\n");
Firebase.setFloat("/SumX1X2",SumX1X2);
Firebase.setFloat("/SumX2Y",SumX2Y); h = dht.readHumidity();
Firebase.setFloat("/SumX1square",SumX1square); t = dht.readTemperature();

int c = Firebase.getInt("/count"); //Firebase.setFloat("/SensorTemperature",t);

//Firebase.setFloat("/SensorHumidity",h);
float b1 = ((SumX2square*SumX1Y)- // Firebase.setFloat("/SensorMoisture",sensor_analog);
(SumX1X2*SumX2Y))/((SumX1square*SumX2square)- sensorMoisture= ((float)(moisture_percentage)/10);
pow(SumX1X2,2)); delay(5000);
if (sensorMoisture<6 && sensorMoisture<KCET0 && PrecipProb<KCET0){
Serial.println("\MotorON");
digitalWrite(4,LOW);
delay(2000);
}
if (sensorMoisture<6 && sensorMoisture<KCET0 && PrecipProb<KCET0){
Serial.println("\MotorON");
digitalWrite(4,LOW);
delay(2000);
}
while(sensorMoisture<7 && sensorMoisture<KCET0 && PrecipProb<KCET0){
sensor_analog = analogRead(A0);
moisture_percentage = ( 100 - ( ((sensor_analog-300)/724.00) * 100 ) );
sensorMoisture= ((float)(moisture_percentage)/10);
Serial.println("Updated Moisture:");
Serial.println(sensorMoisture);
delay(5000);
}
digitalWrite(4,HIGH);
Serial.println("\MOTOR OFF");

//============================================================================================//

Firebase.setInt("/count",c+1);

delay(1800000);
}
//=======================================================================
THANK YOU