You are on page 1of 7

Proceedings of the Second International Conference on Innovative Mechanisms for Industry Applications (ICIMIA 2020)

IEEE Xplore Part Number: CFP20K58-ART; ISBN: 978-1-7281-4167-1

Low Cost IoT based Flood Monitoring System


Using Machine Learning and Neural Networks
Flood Alerting and Rainfall prediction

Dola Sheeba Rani Dr. Jayalakshmi G N Dr. Vishwanath P Baligar


School of Computer Science and School of Computer Science and School of Computer Science and
Engineering Engineering Engineering
KLE Technological University KLE Technological University KLE Technological University
Huballi, Karnataka-580020 Huballi, Karnataka-580020 Huballi, Karnataka- 580020
sheebadjf@gmail.com jaya_gn@kletech.ac.in vpbaligar@kletech.ac.in

Abstract—The term Internet of Things [IoT] refers require speed in change of river. In this system we
to the ever expanding complex system of basic things make use of three rain sensors and three water
that emphasize communication between computing sensors, which are kept at three different locations;
objects, devices and systems by offering connectivity also a raspberry pi with water sensors, rain sensors.
from anyplace and at any time. It is estimated that by This system predicts the flood and alert is produced
the end of 2020, 50 billion devices are said to be for respective authorities while sounding instant
connected. IoT technologies play a crucial role to alarm in nearby villages to transmit message about
encompass many smart applications in real life. On the possible floods using IoT. The system also calculates
other hand, the crosscutting nature of IoT components
the time taken for the flood to reach the desired area.
and systems, introduce new security challenges. IoT
covers advantages for various fields like agriculture,
Water flow sensors measure the water level in an
industry, healthcare, automobiles and home automation water body, whereas rain sensors measure the
for improving and automating various day-to-day intensity of rainfall in millimeters. The entire
activities. Flood is usually caused either by change in the information can be viewed from a remote location,
state of water body or due to the overflow of rivers, providing the required information on IoT gecko
dams, etc. Due to advanced civilization and improved platform.
human life, environment problems also tend to increase.
Rainfall main environment factor to influence the
This paper includes the effective and flexible method for
probability of flood, and a many of machine learning
the detection of flood and alerting system. The most
advanced technologies like machine learning (ML)
models were assessed for flood prediction models.
provide significant boon to the field of technology are The main contribution of the paper presents the
very powerful in monitoring the normal and abnormal contribution of machine learning models. The work
behavioral characters of any machine. The objective of on the innovation of flood prediction models have
this work is to survey on flood issues. Neural networks contributed to minimize the loss of human life,
are most popular, widely used for rainfall forecasting decrease disaster, and reduce destruction of properties
and perform efficiently. and valuables. This paper provides the machine
learning technique, to predict the arrival of floods, by
Keywords—Machine learning, internet of things, considering the rainfall data from previously
neural networks, sensors, embedded system, linear available data to predict the rainfall for next month.
regression, logistic regression, SVM. While building model, it does not require any
meteorological and geographical knowledge about its
I. INTRODUCTION hydraulic and topographic properties. From the
survey, the existing flood forecasting models are:
In the year 2019, there was a heavy rainfall in the radar systems, hydrographic analysis, rain gauge
India that created much loss for 13 states, which network, stream networks, linear statistical models
included loss of at least 200 people and millions of and time series analysis prediction models; these
them were misplaced. The most affected region was
models helped to forecast rainfall in the previous
Karnataka and Maharashtra. Many places such as years. The ML models used in this work for training
coastal areas or the places nearing the river or dams are logistic regression, linear regression, SVM, neural
are also prone to flood disaster. To develop a real networks. Rainfall prediction is considered very
time flood monitoring and alerting system to reduce important, in the aspects of monsoon times. By
the exposure to the flood on real time basis, which
considering the rainfall of previous data, prediction

978-1-7281-4167-1/20/$31.00 ©2020 IEEE 261

Authorized licensed use limited to: UNIVERSITAS GADJAH MADA. Downloaded on September 04,2022 at 12:56:20 UTC from IEEE Xplore. Restrictions apply.
Proceedings of the Second International Conference on Innovative Mechanisms for Industry Applications (ICIMIA 2020)
IEEE Xplore Part Number: CFP20K58-ART; ISBN: 978-1-7281-4167-1

can be done for both short-term and long-term rainfall for predicting water levels for 1-3 hourly basis before
prediction for next consecutive months/ weeks. To flooding.
predict or forecast the rainfall, data is retained from
Indian meteorological department. Two different Rezaeianzadeh; 2014 [9] - by collecting dataset
datasets that consists of average rainfall data from from multiple meteorological stations. Presents many
1951-2000 for every month and each district; second forecasting systems for daily flow prediction using
data is of 1901-2015, which consists average rainfall ANN and multiple linear regressions. Furthermore,
data for each state is used. The performance of the performance of the models is calculated.
trained models were evaluated on mean absolute error Choubin; 2018 [9] - evaluated the accuracy of
and standard deviation. artificial neural networks, but by considering three
common ML modeling tools, of CART, SVM, and
II. LITERATURE SURVEY MLP. The result evaluated that CART performed
well.
Flood issues in India are devastating, to implement
real time flood monitoring systems, many of the Kim and Singh 2013 [10] - presented, three
models have been used so far, with different networks prediction models of generalized regression ANN
for wireless sensor networks, and some energy (GRNNM), Kohonen self-organizing feature maps
efficient models. ANN (KSOFM-NNM), and MLP for flood prediction
and finally evaluated performance of models.
A. A Survey on Flood issues implemented on IoT
Chang, Shen 2014 [11] - presented an advanced
hybrid model, by artificial neural networks for
Ivan Stoianov; 2007 [1]- Presented the use of casting real time flood forecasting model.
PIPENET ( A system based on WSN) that detects,
localize and quantify bursts, leaks and other water Partal and Cigizoglu 2009 [12] - presented a daily
abnormalities such as blockages in the pipelines. forecasting water level, by using WNN and ANN
models.
F. Shebli; 2007 [2] - Presented the method to
calculate the amount of energy consumption by From the survey, the existing system presents
sensors in an network, according to data flow rate, flood issues by interfacing IoT with Wireless sensor
number of nodes and distance between them. Networks; also minimizing the energy consumption
by sensors in an network. Further the survey brings
Jong-uk Lee; 2008 [3] - Presented a real time out the works of implementing the Real-time flood
flood monitoring system with WSN which is monitoring system to ease the flood issues.
deployed in 2 Volcanic islands. This measures river
and weather conditions through Wireless sensors
integrated with different sensors. III. OBJECTIVES
1. To develop the Flood Monitoring System.
Sultanullah Jadoon,; 2011 [4] - Presented efficient
water management system in Pakistan by least cost 2. To interface water level detector and
framework based on sensor network, to manage warning with PC-based system and to send
maintenance of water resources and controlling water warning messages to the intended users.
wastage.
3. To pre-process historical weather data into a
Victor Seal; 2012 [5] - Presents flood forecasting form that is suitable for training and to
model designed using WSN, this model predicts identify the salient variables for training
floods in rivers using the simple calculations to models.
provide real time results and save the lives of people
4. To determine the architecture of ML models
by ringing alarm.
that yield the best predictive performance for
Naveed Ahmad; 2013 [6] - A simulation tool to precipitation.
identify pre and post disaster flood risk analysis by
5. To predict the results by using various
using AdHoc WSN and Presented a study of flood
analysis and prediction using GIS(Geographical machine learning models and to analyse the
Information system). performance of each model.

B. A Survey on Flood forcasting and prediction IV. PROPOSED SYSTEM


using ML Design and implement IoT based flood monitoring
Jimeno-Sáez; 2007 [7] – by applying Artificial and alerting system and calculate the time it would
neural networks for the data set collected from take for the flood to reach them and provides a time
fourteen stream flow gauge stations, presented a flash to people so that they can evacuate accordingly.
flood. Predict the severity of rainfall by machine learning
models.
FJ Chang; 2006 [8] – by collecting data from five
gauges from 1971 to 2001 of water levels has been
used. Successfully provided a water level forecasting,

978-1-7281-4167-1/20/$31.00 ©2020 IEEE 262

Authorized licensed use limited to: UNIVERSITAS GADJAH MADA. Downloaded on September 04,2022 at 12:56:20 UTC from IEEE Xplore. Restrictions apply.
Proceedings of the Second International Conference on Innovative Mechanisms for Industry Applications (ICIMIA 2020)
IEEE Xplore Part Number: CFP20K58-ART; ISBN: 978-1-7281-4167-1

V. BASIC COMPONENTS /TECHNOLOGIES USED of this system includes a flow diagram for rainfall
prediction using ML models.
A. PCB ( printed circuit board)
A. Block diagram of IoT interfacing with IoTGecko:
A printed circuit board is a thin fiber glass that is
used in in different electronic devices such as laptops, Flood monitoring module consists of many hardware
computers, and microcontrollers; where small components like raspberry pi 3, Wi-Fi module, TFT
electrical components like transistor, capacitors, IC’s, display, water sensor, rain drop sensor, resistors,
resistors, diodes etc are connected by imprinted path capacitors, transistors, cables and connectors,
etched on the board. diodes, PCB, LCD, transformer/adapter, push
buttons, switch, IC. Interfacing all these components
on PCB, with raspberry pi, along with three rain drop
B. Water- float sensors
sensors and water level sensors will be placed at
Water- level sensors are used for measuring the different locations. When the water level rises in the
continuous flow of the water or liquid content in a water body, the buzzer is beeped and the system gives
water system. The various applications of these an alert message; with the time remaining to flood in
sensors are used in water tanks, water-level the area.
indicators, water-level alarms etc. The average life
cycle of this sensor is of 15 years. This sensor will
not deteriorate. The circuit of this sensor consists of
GND and VCC. This sensor senses the level of water
and actuates the pump or initiates an alarm by a
buzzer.

C. Rain drop sensors


This is a most easy tool for rainfall detection. For
measuring the intensity of rainfall, this can be used as
sensor. This module consists of 4 pins; GND, VCC,
Digital output, Analogue output; the output can be
connected to rain board or bread board. The
applications of this sensor are used in irrigation
systems, satellite communications, and automobile.
D. Raspberry Pi
Raspberry pi is a low cost, credit card sized
microprocessor, which is plugged into a monitor or
display. This uses a DC power supply, keyboard, and
mouse. It runs on NOOBS operating system. It is
capable of doing all the operations performed by
computer; playing a HD video, computational
operations, play games etc.
Fig 4.1: Block diagram of IoT architecture
E. NOOBS
The development of this project is firstly to choose,
New Out Of Box Software (NOOBS) is an easy the right development board, designing the circuit
installer, an operating system for raspberry pi. Its board, PCB printing, lastly finding the connectivity
open, automatically generated by installing from local issues, the coding is done on python. This platform is
software or from a remote repository. built on a cloud platform, IoT Gecko. The
information sent by the sensors, can be monitored
F. IoT Gecko from any place, just by providing username and
IoT is growing day by day, with more connected password.
device. IoT helps to control more than one object
connect by internet to develop IoT board on own; IoT
Gecko is an open platform with API support for
development boards like Arduino, raspberry pi, and
other microcontroller boards. By setting up the
devices, the system can directly run on the cloud
platform.
VI. ARCHITECTURE
Flood monitoring and alerting system is built by
interfacing IoT hardware with Gecko platform, which
can be monitored by a remote location. The next part

978-1-7281-4167-1/20/$31.00 ©2020 IEEE 263

Authorized licensed use limited to: UNIVERSITAS GADJAH MADA. Downloaded on September 04,2022 at 12:56:20 UTC from IEEE Xplore. Restrictions apply.
Proceedings of the Second International Conference on Innovative Mechanisms for Industry Applications (ICIMIA 2020)
IEEE Xplore Part Number: CFP20K58-ART; ISBN: 978-1-7281-4167-1

VII. RESULTS AND SNAPSHOTS

A. Intefracing hardware with raspberry pi

The above setup shows the real time working of


Fig 4.2: Circuit Diagram of PCB embedded with flood monitoring, this prototype measures the water
hardware components. level in the flood prone area, 3 water sensors and rain
drop sensors measure the intensity of flood.
B. Basic flow of building ML model for rainfall
prediction. B. LCD display steup with PCB components:
To build a rainfall prediction system, meteorological
data, history, many sensing devices etc are required.
The basic traditional water flow data, and other water
gauges.

Fig 4.3: Basic flow diagram of ML rainfall


prediction model
There are three rain drop sensors in the system, which
Finally the models were to undergo training and measures the rainfall in mm/hr, the left description is
testing. about showing the time left out to flood the area.

978-1-7281-4167-1/20/$31.00 ©2020 IEEE 264

Authorized licensed use limited to: UNIVERSITAS GADJAH MADA. Downloaded on September 04,2022 at 12:56:20 UTC from IEEE Xplore. Restrictions apply.
Proceedings of the Second International Conference on Innovative Mechanisms for Industry Applications (ICIMIA 2020)
IEEE Xplore Part Number: CFP20K58-ART; ISBN: 978-1-7281-4167-1

C. Monitoring over IoTgecko: D. Linear Regression model for rainfall prediction

For linear regression the model prediction data is


formatted in the way, that give the rainfall in the last
three months by try to predict the rainfall in the next
consecutive month, and drawing a straight line where
X(Time duration) and Y (Rainfall Measurement).
Testing metrics: used Mean absolute error to train
this model is 40.2467874.

Observations: All means are standard deviation


observations are written, first one represents ground
truth, second one represents predictions.

Readings from the sensors transmitted to the Gecko


cloud. The table shows the results of the sensors and
time taking to flood the area.

978-1-7281-4167-1/20/$31.00 ©2020 IEEE 265

Authorized licensed use limited to: UNIVERSITAS GADJAH MADA. Downloaded on September 04,2022 at 12:56:20 UTC from IEEE Xplore. Restrictions apply.
Proceedings of the Second International Conference on Innovative Mechanisms for Industry Applications (ICIMIA 2020)
IEEE Xplore Part Number: CFP20K58-ART; ISBN: 978-1-7281-4167-1

E. Support Vector Machine. flood, privation of house and office property, streets
causing flood water consists of bacteria and sewage
For SVM the model prediction data is formatted in flow of waste sites and chemical spillage which leads
the way, that give the rainfall in the last three months to a variety of diseases later on. Implementation of
by try to predict the rainfall in the next consecutive flood monitoring and alerting system leads to
month, and drawing a hyperplane to separate the minimization to flood exposure, having time left to
classes and objects. Testing metrics: used Mean evacuate the area, getting time and help to flood
absolute error to train this model is 90.606787. control structures. Various visualizations of data are
observed which helps in implementing the
Observations: All means are standard deviation approaches for prediction. An observation indicates
observations are written, first one represents ground machine learning models won’t work well for
truth, second one represents predictions,Bad prediction of rainfall due to fluctuations in rainfall.
performance by SVR However this method can be used for prediction for
model. any state of India, with the given data. Designed and
implemented the flood monitoring and alerting
system and also helps to predict rainfall for the future
use successfully.
ACKNOWLEDGMENT
The successful completion of any task is
incomplete without complementing those who made it
possible and whose guidance and encouragement
made my effort successful.
I am indebted to my beloved Guide Dr. Jayalakshmi
G N who was constant source of enthusiasm and
whose sincere guidance, valuable suggestions and
F. Artificial Neural Networks benevolent direction where mainly responsible for my
successful completion of the project.. I also thank
Observations indicate that machine learning models
K.L.E Technological University for providing
won’t perform well for prediction models. Therefore
computational support.
by using Convolution neural networks, 1-D is
employed in classifying data, which uses the X axis
REFERENCES
data as time duration and Y axis as rainfall in mm,
[1] Ivan Stoianov, Lama Nachman, Sam Madden. PIPENET: A
and the obtained absolute mean error is 21.8097545 Wireless Sensor Network for Pipeline Monitoring. IPSN’07,
Observations: Neural networks perform better than Vol.3, No.1, pp 264-273, 2007.
ML predicting models. [2] F. Shebli, I. Dayoub and J.M. Rouvaen.Minimizing energy
consumption within wireless sensors networks. Ubiquitous
Computing and Communication Journal, Vol.144, pp. 108-
116, 2007. .
[3] Sultanullah Jadoon, Salman Faiz Solehria, Mubashir
Qayum1. A Proposed Least Cost Framework of Irrigation
Control System Based on Sensor Network for Efficient Water
Management in Pakistan. International Journal of Basic &
Applied Sciences IJBAS-IJENS Vol.978, No.1, pp. 82-86,
2011.
[4] Bashar, A. (2019). Survey on Evolving Deep Learning
Neural Network Architectures. Journal of Artificial
Intelligence, 1(02), 73-82.
[5] Victor Seal, Arnab Raha, Shovan Maity, Souvik Kr Mitra,
Amitava Mukherjee and Mrinal Kanti Naskar. A simple flood
forecasting scheme using wireless sensor networks.
International Journal of Ad hoc, Sensor & Ubiquitous
Computing (IJASUC) Vol.3, No.1, pp 45-60, 2012
[6] Naveed Ahmad, Mureed Hussain, Naveed Riaz, Fazli
Subhani, Sajjad Haider, Khurram.S.Alamgir, Fahad Shinwari.
Flood Prediction and Disaster Risk Analysis using GIS based
Wireless Sensor Networks, A Review.Journal of Basic and
VIII. CONCLUSION. Applied Scientific Research. ISSN 2090-4304, Vol.113,
No.9, pp 632-643, 2013.
Flood is caused when a dam fractures, curtly [7] Jimeno-Sáez, P.; Senent-Aparicio, J.; Pérez-Sánchez, J.;
releasing an immense gallons of water. The result is Pulido-Velazquez, D.; María Cecilia, J. “Estimation of
that a much of the water gets into soil, leading instantaneous peak flow using machine-learning models and
empirical formula in peninsular spain. Water (Switzerland)”
stagnation and causing flooding in the area prone to 2017.
flood. River banks or any water body residing after

978-1-7281-4167-1/20/$31.00 ©2020 IEEE 266

Authorized licensed use limited to: UNIVERSITAS GADJAH MADA. Downloaded on September 04,2022 at 12:56:20 UTC from IEEE Xplore. Restrictions apply.
Proceedings of the Second International Conference on Innovative Mechanisms for Industry Applications (ICIMIA 2020)
IEEE Xplore Part Number: CFP20K58-ART; ISBN: 978-1-7281-4167-1

[8] Chang, F.-J.; Chang, Y.-T. Adaptive neuro-fuzzy inference


system for prediction of water level in reservoir. Advances in
water resources 2006, 29, 1-10.
[9] Shu, C.O., TBMJ. Regional flood frequency analysis at
ungauged sites using the adaptive neuro-fuzzy inference
system. Journal of Hydrology 2008, 349, 31-43.
[10] Rezaeianzadeh, M.; Tabari, H.; Yazdi, A.A.; Isik, S.; Kalin,
L. Flood flow forecasting using ann, anfis and regression
models. Neural Computing and Applications 2014, 25, 25-37.
[11] Tehrany, M.S.; Pradhan, B.; Jebur, M.N. Flood susceptibility
analysis and its verification using a novel ensemble support
vector machine and frequency ratio method. Stochastic
Environmental Research and Risk Assessment 2015, 29,
1149-1165.
[12] Young, C.C.; Liu, W.C.; Wu, M.C. A physically based and
machine learning hybrid approach for accurate rainfall-runoff
modeling during extreme typhoon events. Applied Soft
Computing Journal 2017, 53, 205-216.

978-1-7281-4167-1/20/$31.00 ©2020 IEEE 267

Authorized licensed use limited to: UNIVERSITAS GADJAH MADA. Downloaded on September 04,2022 at 12:56:20 UTC from IEEE Xplore. Restrictions apply.

You might also like