11th IEEE International Conference on Communication Systems and Network Technologies

Prediction and Prevention of Water Pollution

2022 IEEE 11th International Conference on Communication Systems and Network Technologies (CSNT) | 978-1-6654-8038-3/22/$31.00 ©2022 IEEE | DOI: 10.1109/CSNT54456.2022.9787664

by Industries using IoT based Monitoring

System
Abha Porwal* Sambit Kumar Mishra* Arya Kela
Electronics and Telecommunication Electronics and Telecommunication Electronics and Telecommunication
Engineering Engineering Engineering
Shri G.S. Institute of Technology Shri G.S. Institute of Technology and Shri G.S. Institute of Technology and
and Science Science Science
Indore, India Indore, India Indore, India
[email protected] [email protected] [email protected]

Gargi Singh Manish Panchal Anjana Jain

Electronics and Telecommunication Electronics and Telecommunication Electronics and Telecommunication
Engineering Engineering Engineering
Shri G.S. Institute of Technology Shri G.S. Institute of Technology and Shri G.S. Institute of Technology and
and Science Science Science
Indore, India Indore, India Indore, India
[email protected] [email protected] [email protected]

Abstract—The ever-increasing water pollution level due to availability of many anti-pollution laws, indicates the lack
industries, pose a serious threat to the masses and once of the mechanism for holding a potential source
detected, it requires a significant amount of time for the responsible. This calls for an immediate attention to the
remedial measures. The proposed IoT solution discusses a major pollution contributing sources and assessment of
device, capable of determining the water pollution level caused their contribution to the overall pollution. Among the
by an industry, near water bodies. The system would also be contributing factors to water pollution, the focus has been
able to predict when the pollution level tends to cross the kept on the industrial belt near the river as they are the
threshold value set by C.P.C.B. India in future. This major contributing factors as per the existing research. The
information would eventually help in taking proactive use of LoRa in IoT devices for wireless data transmission is
remedial measures.The proposed device consists of four major not uncommon. LoRa's performance is evaluated using
sensors: pH sensor, turbidity sensor, flow sensor, and
three key parameters: code rate, spreading factor, and
temperature sensor. These sensors would measure the water-
parameters in real-time and send that data to Google Firebase bandwidth. Lora has five forward error correcting code
(database) either through the WiFi module or, through the rates, allowing for the recovery of bits of information that
LoRa transmitter installed in the proposed device. Through have been corrupted by interference [1].
LoRaWAN, all of these devices, which will be put at various The proposed system has been created keeping in mind
industry-wastewater-outlets, will be wirelessly connected to a all the aforementioned points in mind. There has always
central hub. The goal is to use an optimized algorithm to make been a need to identify the amount of pollution level in the
the system more energy efficient. The database would retain water in real-time and consequently, hold an industry
real-time data for two purposes: displaying it in the mobile responsible for the same, backed with proof. Once the
app and using it in LSTM time-series forecasting (Machine major contributing sources have been identified, necessary
Learning model) to predict when pollution levels could surpass action can be undertaken to put a check on the wastewater
the threshold.
release into the river.
Keywords—IoT, water pollution, LoRaWAN, turbidity, pH, Subsequent sections describe the rest of the paper as
Machine Learning, time-series forecasting, LSTM, mobile
follows. Section II gives the background for the work and
application, real-time, Google Firebase
benchmarks required. Working flowchart and prototype
I. INTRODUCTION for water pollution monitoring is explained through
section III and section IV. Results are described in section
The major problem that wreaks havoc in the healthy V. Next section mentions the application of the project and
lives of people nowadays is the unavailability of clean finally, the paper is concluded through the last section.
water for consumption. The water streams like rivers, lakes
are getting polluted at an ever-increasing alarming rate.
This indiscriminate pollution of the water bodies despite the

978-1-6654-8038-3/22/$31.00 ©2022 IEEE 504

DOI: 10.1109/CSNT.2022.88
Authorized licensed use limited to: Odisha University of Technology and Research. Downloaded on August 06,2025 at 14:59:56 UTC from IEEE Xplore. Restrictions apply.
 II. BACKGROUND energy-efficient data transmission mechanism, with the
A. Literature Review energy being captured using solar panels to extend the
Extensive research has been carried out on the sensor's lifetime. N S Kamaruidzaman et al., examines the
detection of pollutants in the water bodies. There are use of IoT in water monitoring systems in particular [9].
effective methods to determine the pollution level in water The paper proposes the delivery of real-time monitoring
bodies andthe usage of IoT to transmit the data. But, the information, allowing consumers to anticipate the actions
methods to identify the industries contributing to the that may be required. Some data can be sent and stored in
pollution majorly, is still a domain which has a potential centralized or cloud distributed repositories as virtual
of extensive research. datasets via GSM and wifi module.
The paper presented by S.Geetha and S.Gouthami et al. Dr. Nageshwarao Moparth et al., proposes a water
[2], highlights the advancements in the field of water quality monitoring system to find the quality of the water
quality monitoring using IoT. The user is alerted using and send it to authorities [10]. The device measures the pH
the data received from the sensors. The whole system has of the water and and using an Arduino board and a GSM
been divided into three subsystems: Data module the data is sent. Parag Warungase et al., proposed
Management Subsystem, Data Transmission to measure turbidity, pH and temperature of the water in the
Subsystem, Data Collection Subsystem. The paper also paper [11]. To transfer the data, the paper suggested the use
discusses some of the common applications of such a of zigbee technology. It has realized the automation of
system, which includes household water quality monitoring intelligence of data analyzing and
consumption, aquaculture etc. networking of information transferring.
Jayati Bhatt et al., shows the IoT based water quality Vennam Madhavireddy et al., describes the usage of
monitoring system in real-time [3]. The data from the controller ARM with a Wireless Fidelity Protocol to
sensors is processed in the Raspberry Pi microcontroller and interface the values received from the sensors [12]. The
then transmitted to the server using the Zigbee Protocol. user is notified through a buzzer in case the sensor values
The paper also provides a comparison between various exceed the threshold limit. The paper also discusses the use
transmission protocols. An insight into the wireless network of a CO2 sensor. Lambrou et al. [13] explored the design
implementation in water-monitoring has been discussed. and execution of a portable, transportable, cost-effective,
Kartik Maheshwari et al., discusses the optimization of and dependable water level management system. Two radio
water use. The robustness and the cost-effectivity of the frequency (RF) transceivers and a transmitter are put on the
system have been given importance [4]. The ThingSpeak tank and sump to test the water quality. The radio frequency
dashboard has been used for online monitoring purposes. transceivers are utilized for wireless connection with the
This paper also discusses the use of SMS services to alert internet. The system is entirely programmed by the user
the users if the sensor readings exceed the threshold limits. using a microcontroller.
Ajith Jerom et al.[5], advocates the importance of the Sathish Pasika et al. attempt to describe a system that
regular monitoring of water quality using IoT. NodeMCU comprises multiple sensors that detect different
has been selected as the microcontroller for the system and characteristics such as pH value, turbidity in the water, tank
the sensor data is transmitted to the cloud servers using the level, temperature, and humidity of the atmosphere.
inbuilt WiFi module of the NodeMCU. The paper Furthermore, the Microcontroller interfaced with these
emphasizes the need to process the time-sensitive data sensors, and further processing was carried out on the
immediately before storing in the cloud and also to avoid Personal Computer (PC). For water quality monitoring, the
network clutter. The sensors and the transmitters are collected data is transferred to the cloud using an IoT-based
attached on a buoy which keeps the hardware afloat for ThinkSpeak application [14]. Another paper proposes a
transmission and the sensors submerged for the data. system for large aquatic areas. It focuses on Materials and
Muhammad Bilal et al., discusses the aspect of methods for monitoring the water bodies in real time using
measuring the water quality through image processing [6]. WSM. The paper talks about the system design flow using
The methods used to implement it include Image Variance web servers [15].
(IV) and Threshold Segmentation with Path Discretization There are various other papers which discuss the
(TSPD). The paper also discusses the other existing different methods to measure water pollution. Also a lot of
methodsto determine the water quality. Farman Ullah Jan et work has been done on transmitting data wirelessly using
al., discusses the importance of water quality monitoring LoraWAN.
for household purposes [7]. This paper also provides brief
information on various water quality monitoring methods. B. C.P.C.B. Benchmarks
Appropriate comparisons have been made for those The water quality is measured by two factors i.e.
methods. physical properties and chemical properties. Although the
The paper described by Ferdinand et al.[8], discusses the majority of the properties can’t be measured with the help
general needs of a WQM system including high data of sensors, we can measure the ones like pH, turbidity, flow
precision, data correctness, timely reporting, data and temperature to have insights of the level of water
accessibility, and completeness. This study proposes an pollution.

505

Authorized licensed use limited to: Odisha University of Technology and Research. Downloaded on August 06,2025 at 14:59:56 UTC from IEEE Xplore. Restrictions apply.
 Table 1. shows the threshold values of various water- architecture of the device which has been proposed.
pollutants set by the C.P.C.B. for the water usage in various
household and industrial purposes. This data has been used
in the implementation of the proposed device.
TABLE 1. Threshold of water pollutants set by C.P.C.B. [16]

Figure 1. Block diagram of the implementation

The block diagram also depicts how multiple devices can

be integrated with the LoRa module for integrated
III. METHODOLOGY
communication with our designated database in the remote
The proposed system will act as a solution to monitor server. The data can later be processed, and the concerned
the water pollution by industries in such a way that a authorities can be alerted if the pollution level exceeds the
particular industry can be held responsible for the threshold or is expected to exceed the threshold in the near
increasing level of water pollution. The system proposed
future.
has the following features:
● Measure pH and inform whether the water is acidic D. Algorithmic Flowchart
or basic and also the extent of acidity and basicity. The algorithm that is finalized for the implementation
● Measure the turbidity of water to find out the begins with the initialization of a variable “pFlow” which
quality. would be initialized with the maximum normalized value of
● Measure the flow of water to note the volume of the flow-sensor. Another variable named ‘flow’ is declared
water released per day. which would store the current value of the flow sensor. Then
● Measure the temperature of water. “pFlow” would be assigned the minimum value amongst
On the basis of these points as well as the threshold “flow” and “pFlow”. To avoid reaching the exception case
decided by CPCB, the proposed solution will also inform of value 0 of “flow”, the value of “pFlow” has been kept
the level of water pollution caused by the industry. above a certain practical threshold value.
Machine learning models are also going to be trained so Now current flow is measured again. If the value of
as to predict the future pollution by the industry based on “flow” is greater than “pFlow” then it means that the
the past data so that if the model predicts that after x days, industry has discharged wastewater. This is the condition
the pollution level may cross the threshold, then it will warn for triggering the proposed device to activate all the
that industry by sending them an alert. Also, if it crosses the remaining idle sensors.
threshold, it will directly send a red alert to the concerned Once all the sensors in the device have been activated,
authorities. the readings of all the sensors are going to be compared
The paper proposes the use of LoRaWAN technology withthe thresholds and an alert will be sent accordingly.
so that various monitoring devices in the industrial belts can
communicate with each other in order to regulate and
monitor the pollution level and also to determine whether
there is some industry which is releasing wastewater
without informing the regulatory bodies.
C. Block diagram of the proposed system
The block diagram in Figure 1 shows the internal

506

Authorized licensed use limited to: Odisha University of Technology and Research. Downloaded on August 06,2025 at 14:59:56 UTC from IEEE Xplore. Restrictions apply.
 2) Power Calculation: As we are using Li-Ion
Battery (2500 mAh), the power calculation is done with
some suitable assumption based on Industrial standards.
The equation used for power calculation is:
Power required = total power of
sensors
Battery power = (no. of days) * (Power required) The
total battery life comes out to be approximately 3 months.
This has been possible because of the algorithmic
optimization of the activation time of the sensors. In
comparison with the original battery life, the optimization
ofpower has increased the average battery life 20 times.
3) Mobile-App for real-time data viewing: To view
the sensor data in real-time a mobile application has been
created which is connected to Google Firebase. The data in
the application is updated as soon as Firebase receives a
change in the parameter value.
4) ML algorithm implementation: To predict the
future level of pollution in order to alert the industries in
advance using time series forecasting - LSTM (Long Short
Term Memory). LSTM has been preferred because of the
Figure 2. Flowchart of the operation
availability of memory cell, and peep-hole connections. The
LSTM allows control of the large number of weights of the
network, thus making it suitable for random data.
IV. PROTOTYPING Long Short Term Memory i.e. LSTM is a kind of RNN
The microcontroller used for the implementation is and each cell of LSTM consists of 4 gates namely - Input
NodeMCU (ESP8266) because it is cost-efficient and Modulation Gate, Input Gate, Forget Gate and Output Gate
comes with an in-built WiFi Module. All the four sensors - having the cell structure shown in figure 3 -
Flow Sensor (YF-S201), Turbidity Sensor (SKU-62828),
pH Sensor (SKU-235871), Temperature Sensor
(DS18B20) - are connected with the NodeMCU. The
sensors are powered by the Li-Ion Battery (2500 mAh). The
assembly of the device has been depicted in Figure 4. The
sensor read data is being transmitted to the Google Firebase
database through the in-built WiFi module (in the case of
availability of WiFi signal) or through the LoRa Module (in
the case of unavailability of WiFi signal i.e. when the
device is installed at a remote location). There is a central
secure LoRa receiver module which receives the data from Figure 3. LSTM Cell structure
the remote devices having LoRa transmitter module
installed in them. The size of the dataset used to train the model is 1000. It is
The Flow Sensor will always remain active during the further subdivided into two groups - training set (650 data
operation of the device and will trigger the remaining points) and validation set (350 data points).
sensors whenever the device detects a wastewater The model consists of three layers having 200 nodes
discharge. This technique of selective activation of sensors having ReLu as an activation function using Adams as an
ensures the battery optimization. Once all the sensors have optimizer. Present model is trained using 100 epochs and a
been activated, they start recording the respective data and batch size of 128. All these parameters can further be
send the same to the remote Firebase database through tweaked in order to achieve a better performance.
LoRaWAN.
Once the data is received in the database, it is sent to the V. RESULTS
mobile app for the real-time viewing and it is also utilized While testing the device, the surrounding conditions
for predictive analysis using time-series LSTM Algorithm. were kept as much similar to the actual scenario as possible.
1) Sensor Integration with NodeMCU: The circuit The following results were obtained after careful
diagram of the device is shown in Figure 4. It includes the experiments under multiple experimental environments.
interfacing of all the 4 sensors with the NodeMCU. The The components have been closed packed together
sensors are connected with the help of a 3x8 mux IC, due to during the assembly to enhance the compactness of the
less number of analog pins in the microcontroller. device. The water-proofing of the device to sustain in

507

Authorized licensed use limited to: Odisha University of Technology and Research. Downloaded on August 06,2025 at 14:59:56 UTC from IEEE Xplore. Restrictions apply.
 various water conditions has been taken care of.
The performance of the model is being judged on the
basis of RMSE that is Root mean square error. RMSE is the
standard deviation of the prediction errors. RMSE values in
the range 0.2 to 0.5 indicate that the model can reasonably
predict the data. The implemented LSTM model has been
tested to have a RMSE (Root mean square error) of 0.657.
Table 2. shows the accuracies obtained in prediction of
values for respective sensors.
TABLE 2. Accuracy of predicted values

Property MSE Accuracy(%)

pH 1.25 96.60
Figure 6. Real v/s Predicted value of Turbidity Sensor
Temperature (C) 1.59 91.04
The mobile application provides real-time insights into
Turbidity (NTU) 1.47 90.22 the data sent by each device in the device-network. Real
time alerts are notified in the phone itself and necessary
The model can be further optimized with the availability actions can be taken if needed.
of field data. The actual and predicted values of different VI. APPLICATION
parameters like pH (Figure 4.), Temperature (Figure 5.) and
Turbidity (Figure 6.) are as follows: The IoT device proposed as the solution to the problem
statement is majorly targeted at the industrial wastewater
releasing industries near the river bodies. The detailed
applications are mentioned as follows:
A. Industrial belt near the river like Yamuna: We plan to
install this device near the opening of the industrial
wastewater release pipeline. The device will be in
standby state always and it will get triggered once the
wastewater is released by the concerned industry. If
the data received by our device senses a spike in the
pollution level, then it will alert the concerned
authorities.
B. Useful in monitoring the water pollution in real-time
in any industry: This device can be installed in any
industry which can be considered as a potential
contributor to the rising water pollution. In this way,
the concerned authorities can keep a track and assign
Figure 4. Real v/s Predicted value of pH Sensor responsibility to the particular industry with full
confidence.
C. Unlawful wastewater release in a particular water
body can be detected efficiently: Installing this device
in static water bodies near industrial areas can also be
helpful to determine which particular industry is
causing a spike in the pollution unlawfully.
VII. CONCLUSION
The significance of this project may be seen by looking at
the ever-increasing amounts of water contamination caused
by industrial effluents. Water pollution may be significantly
reduced if we can identify the specific industries that are
contributing to rising pollution levels and report them to the
appropriate authorities. Although the implementation of the
prediction algorithm has been tested to have an RMSE value
Figure 5. Real v/s Predicted value of Temperature Sensor
of 0.657, there might be some false-positives in the real-

508

Authorized licensed use limited to: Odisha University of Technology and Research. Downloaded on August 06,2025 at 14:59:56 UTC from IEEE Xplore. Restrictions apply.
 world implementation. This shortcoming can be countered (MACS). IEEE, 2018.
[7] Jan, Farmanullah, Nasro Min-Allah, and Dilek Düştegör. “Iot
once real-time data is received and appropriate based smart water quality monitoring: Recent techniques, trends
modifications to the learning algorithm are implemented. and challenges for domestic applications.” Water 13.13 (2021):
This device is expected to work well within the standard 1729.
industrial environment. This device can be further modified [8] Adu-Manu, Kofi Sarpong, et al. “Smart river monitoring using
wireless sensor networks.” Wireless Communications and Mobile
externally in order to scale the application of this device in Computing 2020 (2020).
drastic conditions. The algorithm followed to determine the [9] Kamaruidzaman, N. S., and Siti Nazahiyah Rahmat. “Water
responsible industry has included all the threshold standards monitoring system embedded with internet of things (IoT) device:
set by the governmental agency C.P.C.B. a review.” IOP Conference Series: Earth and Environmental
Science. Vol. 498. No. 1. IOP Publishing, 2020.
ACKNOWLEDGMENT [10] Moparthi, Nageswara Rao, Ch Mukesh, and P. Vidya Sagar.
“Water quality monitoring system using IoT.” 2018 Fourth
This work is supported by Shri G.S Institute of International Conference on Advances in Electrical, Electronics,
Technology and Science. We would like to show our Information, Communication and Bioinformatics (AEEICB).
gratitude to our faculties for their valuable feedback. We IEEE, 2018.
would also like to extend our heartfelt regards to our family [11] Warungase, Parag, et al. “IOT Based Water Monitoring System.”
[12] Madhavireddy, Vennam, & B. Koteswarrao. “Smart Water
and friends for their constant support throughout. Quality Monitoring System Using Iot Technology.” International
Journal of Engineering & Technology [Online], 7.4.36 (2018):
REFERENCES 636-639. Web. 18 Feb. 2022.
[1] Noreen, Umber, Ahcène Bounceur, and Laurent Clavier. “A study [13] Lambrou, Theofanis P., et al. “A low-cost sensor network for
of LoRa low power and wide area network technology.” 2017 real-time monitoring and contamination detection in drinking
International Conference on Advanced Technologies for Signal water distribution systems.” IEEE sensors journal 14.8 (2014):
and Image Processing (ATSIP). IEEE, 2017. 2765-2772.
[2] Geetha, S., and S. J. S. W. Gouthami. “Internet of Things enabled [14] Pasika, Sathish, and Sai Teja Gandla. “Smart water quality
real time water quality monitoring system.” Smart Water 2.1 monitoring system with cost-effective using IoT.” Heliyon 6.7
(2016): 1-19. (2020): e04096.
[15] Dinesh Sharma, GS Tomar, “Energy Efficient Multitier Random
[3] Sabari, M., et al. “Water Quality Monitoring System Based on
DEC Routing Protocol for WSN: in Agriculture”, Wireless
IoT.” 5th International Conference on Devices, Circuits and
Personal Communication, Vol.120, No.1, pp.727-747, 4th May
Systems (ICDCS). IEEE, 2020.
2021.
[4] Maheshwari, Kartik, and Adrija Chakraborty. “Water Quality
[16] Demetillo, Alexander T., Michelle V. Japitana, and Evelyn B.
Monitoring System Implemented With IoT.” International Journal
Taboada. “A system for monitoring water quality in a large
9.7(2021).
aquatic area using wireless sensor network technology.”
[5] Manimegalai, R. “An IoT based smart water quality monitoring
Sustainable Environment Research 29.1 (2019).
system using cloud.” 2020 International conference on emerging
[17] Cpcb.nic.in. 2022. CPCB | Central Pollution Control Board.
trends in information technology and engineering (ic-ETITE).
[online] Available at: <https://cpcb.nic.in/water-quality-criteria/>
IEEE, 2020.
[6] Bilal, Muhammad, et al. “A study on detection and monitoring of
water quality and flow.” 2018 12th International Conference on
Mathematics, Actuarial Science, Computer Science and Statistics

509

Authorized licensed use limited to: Odisha University of Technology and Research. Downloaded on August 06,2025 at 14:59:56 UTC from IEEE Xplore. Restrictions apply.