1st International WDSA / CCWI 2018 Joint Conference, Kingston, Ontario, Canada – July 23-25, 2018

IoT ENABLED MONITORING AND CONTROL OF WATER

DISTRIBUTION NETWORK
Saravanan Chinnusamy1, Prasanna Mohandoss2, Partha Paul3, Rohit R,
N Murali, S Murty Bhallamudi4, Shankar Narasimhan5, Sridharakumar Narasimhan6
1,2,5,6
Department of Chemical Engineering, Indian Institute of Technology Madras, Chennai, India
4
Department of Civil Engineering, Indian Institute of Technology Madras, Chennai, India
3
Department of Electrical Engineering, Indian Institute of Technology Madras, Chennai, India
sridharkrn@iitm.ac.in

ABSTRACT
Water supply in most of the Indian cities is intermittent with poor levels of service. The poor service
levels can be attributed to constraints on available water and resources, limited instrumentation,
improper operation of the system and poor network maintenance. To monitor and control these
geographically distributed networks, sensors should be placed in appropriate locations and manual
valves should be automated and appropriate communication between the devices enabled. The
proposed method involves real time water level monitoring of storage reservoirs in Water
Distribution Networks (WDN) and remote actuation of valves using Internet of Things (IoT) enabled
devices. A low power wireless sensor and actuator network is developed for monitoring and control
of water distribution networks and deployed in the IIT Madras water distribution network. The
network consists of low cost water level measurement module electrically actuated valves which
function as remote and relay nodes and gateway nodes. LoRa is used to communicate between the
nodes. A Raspberry Pi based gateway is developed which collects the level data from all remote and
relay nodes. The gateway sends the data to local server and also sends control signals to the
actuator. Grafana is used for data visualization and also serves as a human machine interface for
manipulating the valve remotely.

Keywords: water distribution network (WDN), Internet of Things (IoT), LoRa.

1. INTRODUCTION
1.1 IoT in WDN monitoring
Drinking water distribution systems carry potable water from water sources to consumers through
complex pipe networks. Very few Indian cities receive continuous water supply while it is
intermittently supplied in most other cities with an average water supply of 1.3 hours per day [1].
Service reservoirs are used as intermediate storage devices to smooth and meet demands. Most
households also own local storage tanks to further mitigate the variability of water supply. Though
the water distribution network (WDN) is initially designed to supply water continuously, rapid
population growth and unplanned expansion of network results in intermittent supply. Poor service
is further compounded by lack of instrumentation, improper operation and poor network
maintenance. Continuous monitoring of WDN parameters such as, pressure, flow, water quality and
water levels in storage reservoirs can improve the quality of service by balancing supply and
demand and reduce the Non-Revenue Water (NRW). The proposed method involves real time water
level monitoring of reservoirs in WDNs and remote actuation of valves using Internet of Things
(IoT) enabled devices. IoT in WDN is used to collect crucial data about the system for monitoring,
scheduling and equitable supply of water to consumers.
1st International WDSA / CCWI 2018 Joint Conference, Kingston, Ontario, Canada – July 23-25, 2018

1.2 Related work

Recently smart WDN monitoring using Internet of things (IoT) has received attention from research
community recently. The deployment of IoT devices in water distribution network monitoring is
limited by the technical challenges such as long communication distance and cost. Mobile
communication (GSM/GPRS,3G/3G) is still expensive for monitoring WDN and not energy
efficient for remote monitoring devices which are resource and power constrained. Low power,
short range radios such as IEEE 80215.4 with multi-hop routing have been used. However, it
increases the complexity of the network. PipeNet [2] uses wireless sensor network (WSN) for
monitoring underground pipe lines. The authors also discussed about monitoring water quality and
sewage water collector level. WaterWise [3] utilized WSN for real time monitoring of WDN in
Singapore. Low cost WSN is used for remote detection of leaks and pipe line bursts. MISE-PIPE
[4] also used WSN for underground pipe line monitoring.
Recently developed low power wide area network (LPWAN) based on Sigfox and SemtechTM can
provide long range wireless communication with less complexity and cost. LoRa provides better
connectivity and long-distance communication can be achieved without relay nodes. LoRa based
communication can achieve 20 KM or higher in range with line of sight (LOS) and about 2 KM
without LOS [5,6]. It uses unlicensed spectrum and can be deployed in a given area without any
additional subscription, whereas Sigfox is entirely operator managed. Adige [7] a smart water
network monitoring system using long range wireless communication technology LoRa was
proposed. In [8], the authors used LoRa based communication for wireless water meter reading
systems. A low-cost solution for remote monitoring using LoRa and open source IoT platforms have
been developed in [9]. Although several works use LoRa for monitoring applications, none of them
have deployed it for control of valves and pumps. In the proposed method LoRa is used for both
monitoring water levels and for remote operation of valves.
2. SYSTEM ARCHITECTURE
In this work, we focus on a specific class of water distribution networks having the following
characteristics. Water from a source is supplied to overhead and underground storage reservoirs
(OHSR and UHSR respectively), from which water is distributed (by gravity or pumping
respectively) to end consumers. The inlet pipes to the reservoirs are fitted with valves which are
opened and closed in a prescribed schedule to manage the water supply. This is typical of campus,
rural, regional and urban water supply networks in India. It has been demonstrated that pumps and
the inlet valves can be manipulated in an optimal manner and the supply scheduled appropriately to
minimize energy consumption or ensure equitable supply [10].
WDNs are geographically distributed. Monitoring and control of these networks requires
instrumentation, automation of manual valves and proper communication to cover large distances.
The proposed network consists of remote/actuator nodes, relay nodes, gateway nodes, web server
and end devices as shown in figure 1. The remote nodes are installed in storage reservoirs and
electrically actuated ON/OFF valves. It is possible to connect all OHSR in star topology, but the
problem is connecting UGSR to gateway. LoRa uses line of sight for communication and
connecting UGSR to the gateway is difficult. There is no line of sight between UGSR nodes and
gateway. Hence nearby OHSR nodes function as repeater nodes (Relay Node). A relay node not
only transmits its own data, but also sends the data of other remote nodes to the gateway. The
gateway collects the level data from all remote nodes, processes the data and sends the data to
local/cloud server. It also sends the control signal to the actuator. GSM is used for redundancy. Web
and Mobile API is developed to monitor the water level and to control the actuator.
1st International WDSA / CCWI 2018 Joint Conference, Kingston, Ontario, Canada – July 23-25, 2018

Figure 1. System Architecture.

3. HARDWARE DESIGN
3.1 Remote Node
Water level in the reservoirs is usually measured by using float & board level gauge and dip sticks.
The level gauge values are not accurate and require constant human effort. Improper readings can
result in insufficient storage or reservoir overflow. Remote nodes are installed in the OHSR/UGSR
to monitor water levels. It consists of Arduino MCU, LoRa module (AI Thinkers Ra-02 SX1278),
Ultrasonic sensor, GSM Modem. Arduino collects the water level in the reservoir every minute
from ultrasonic sensor and sends to gateway/relay node using LoRa. Arduino Nano MCU is used,
due to its small form factor and low power consumption.

Figure 2. Block diagram of Remote node. Figure 3. Prototype of Remote node.

MaxBotix weather proof ultrasonic sensor (MB7060 XL-MaxSonar-WR) is used to measure water
level in the reservoirs. The sensor and electronic devices are enclosed in IP65 box. GSM (Sim 900A
module) is used as a redundant system. Figure 2 shows the block diagram of the developed module.
1st International WDSA / CCWI 2018 Joint Conference, Kingston, Ontario, Canada – July 23-25, 2018

Separate MCU is used for both LoRa and GSM. LoRa is connected to Arduino using serial
peripheral interface (SPI). Power supply is designed to supply 5V to Arduino, ultrasonic sensor and
GSM. Power for LoRa is taken from Arduino 3.3V output. Remote nodes take reading for every
minute and send to gateway. Lithium ion batteries are used to power remote nodes which do not
have access to electrical mains. Considering three 2200 mAh batteries with battery utilization of
75% and GSM in receive mode, the system can work approximately for 10 days. Solar panels are
used as a power source to charge lithium ion batteries. 20W/12V solar panel is used to charge the
batteries in two stages, Linear voltage regulator followed by charge controller. The system can work
continuously without changing the batteries frequently. Remote nodes are duty cycled to wake up
every minute to save power.
3.2 Actuator Node
In intermittently operated water networks, supply to different parts of the network are regulated
using valves according to heuristically determined schedules. Currently, in many networks, sluice
valves are operated manually by using long T pipes. It is a labor-intensive operation, e.g., more than
45 turns are required to open/close a valve on a 10” line. Moreover, complex operation and
schedules cannot be implemented using manual operation alone. Hence, the existing sluice valve is
retrofitted with an electrical actuator to ease the operation and to enable complex supply schedules.
The block diagram of the actuator node is shown in Figure 4. The actuator has local status
indication (open/close) and button to operate the valve locally. It also has remote control option, i.e.,
it can be opened/closed externally using relay or PLC.
As shown in Figure 5, the electric actuator is fixed on top of the valve using supports and a long
stem is used to transfer the torque from actuator to valve. Potential free contacts are provided to
read actuator status. The remote node reads the actuator status including local/remote mode status,
Fully open, Fully close. This is read every five seconds and transmitted to the gateway using LoRa.
Communication between actuator node and gateway is half duplex. After sending actuator status to
gateway, it waits for short period of time to get control command from the gateway. Based on the
command received, Arduino can energize or de-energize the relay to open/close the actuator. These
nodes are powered from actuator power supply, so batteries are not used. Power supply is designed
to supply 5V to Arduino and GSM. To avoid single point failure, Separate MCUs are used for both
LoRa and GSM. Actuator status and controls are connected to both MCUs. Both LoRa and GSM
are initialized during starting.

Figure 4. Block diagram of Actuator node. Figure 5. Electrical Actuator

1st International WDSA / CCWI 2018 Joint Conference, Kingston, Ontario, Canada – July 23-25, 2018

3.3 Gateway Node

The gateway node collects the data from remote nodes and actuator nodes, processes the data and
sends to centralized server. Commercial LoRaWAN gateways can listen to many LoRa channels
simultaneously. Semtech SX1301 concentrator is typically used and is capable of scanning 8
channels simultaneously. The cost of such SX1301 based gateways is high (~1000 USD). We use a
single channel gateway developed using SX1278 based chips. Raspberry PI is used for data
processing and internet connectivity. The cost of raspberry Pi based gateway is low (~75 USD).
SX1278 based LoRa Modem is connected to Raspberry Pi using serial Peripheral interface (SPI).
Raspberry Pi is connected to centralized server using local area network (LAN). The gateway
pushes the data to central server whenever it receives data from any of the remote/actuator node. It
receives command from the centralized server and transmits to the actuator node to open/close the
actuator. Gateway LoRa is always in receive mode. Python is used for programming Raspberry Pi.
4. COMMUNICATION NETWORK
4.1 LoRa Communication
LoRa is a long-range communication technology developed by Semtech. It is designed for low
power application and operates in ISM band. It uses spread spectrum approach to increase the range
and robustness by increasing receiver sensitivity. LoRaWAN is a low power, wide area network
protocol, designed to allow battery operated devices to communicate with internet connected
application. LoRaWAN is typically deployed in a star topology. Gateways are connected to central
server via standard IP connection. Actuator nodes use single hop wireless communication to the
gateway whereas remote nodes use multi hop communication (it can also act as a relay node to
forward other nodes data). Single gateway covers more than 10 Km in range and cost effective and
hence is an ideal choice for monitoring geographically distributed WDN.
Remote nodes are programmed to work as class A devices to save power. Remote nodes uplink
communication consists of one slot and downlink consists of two shorter slots. Uplink is scheduled
by remote nodes for every minute. Actuator nodes are programmed to act as class C devices. These
devices are always in receive mode except when they are transmitting data to gateway. Class C
devices have maximum downlink slots, require more power and have least latency for
communication between end device and gateway.
4.2 Redundancy in communication
LoRa is the primary communication between all nodes to gateway. Gateway collects remote nodes
level data for every minute, time stamps each data and updates it into the database. Remote nodes,
actuator node and gateway are equipped with GSM to act as a redundant system. In case of failure
of LoRa and consequently if the data is not updated in the database after two minutes, GSM modem
in the gateway sends text message to corresponding remote node GSM modem to send the data
directly to the server. This process is repeated until the LoRa communication is restored. GSM
modem in remote node is always in receive mode. It is also programmed to manually pull the level
data, actuator status and to control the actuator at any time from mobile phone.
5. SOFTWARE DESIGN
We have used modified SX1272 libraries developed in [9] for programming LoRa. It supports
SX1278 type chips without any modification, we have used LoRa based on SX1278 type chips in
gateway, remote nodes and actuator nodes. LoRa is interfaced to gateway and Arduino MCU using
SPI. SX1272/76 library use modified header as shown in Figure 5, packet type field is introduced to
differentiate data and acknowledgement. For example, 0001 represents data and 0010 represents
1st International WDSA / CCWI 2018 Joint Conference, Kingston, Ontario, Canada – July 23-25, 2018

acknowledgement. The library is simple and can be modified according to our need (more address
space, longer sequence number etc.). By default, LoRa in both actuator and end nodes operate in
mode 1(low bandwidth 125 Mhz) to achieve longer distance. When the gateway receives a packet,
it prints everything that it receives from the end-devices such as the source address, sequence
number, payload length, Signal to Noise ratio (SNR) and the Received packet strength index
(RSSI). Python is used for data processing. It pushes the data to central server and receives
command to open/close the actuator.

Figure 5. LoRa packet structure. Figure 6. Webpage to control Actuator.

The data received by the central server is stored in a time series data base. An open source database
InfluxDB is used. It is optimized for fast, high available storage and retrieval of time series data.
Queries are more responsive than SQL, which improves the interface usability. Grafana, an open
source general purpose graph composer, run as web application is used for data visualisation. PHP
is used to create an additional webpage in Grafana to monitor and control the actuator node as
shown in Figure 6.
6. CAMPUS WATER DISTRIBUTION NETWORK
The described network is deployed in the IIT Madras campus water distribution network.

Figure 7. Schematic of campus WDN Figure 8. Network topology

1st International WDSA / CCWI 2018 Joint Conference, Kingston, Ontario, Canada – July 23-25, 2018

The overall network structure is shown in Figure 7. Water is received from Chennai Metropolitan
Water Supply and Sewerage Board and is stored in UGSR1, from there water is pumped to different
OHSRs and UGSR. Distances between the tanks, valves and reservoirs being large, it is difficult to
manually monitor and control the WDN. The network is currently operated manually, and tanks are
filled according to a pre-determined schedule. Remote nodes are installed in OHSR1, OHSR2,
OHSR5, UGSR1 and UGSR4. Inlet valve to OHSR2 is retrofitted with an electrical actuator. Node
3 and node 4 are programmed to function as a relay node to relay OHSR 3 and UGSR 4 data to
gateway. The resulting network topology is tree type as shown in Figure 8. The numbers represent
the node address of each remote/actuator node.
7. RESULTS
Remote nodes send level data for every minute to gateway and the same is updated in the database,
Grafana is used to visualize the current and time series data as shown in Figure 9. Python script in
the gateway checks the status of data received from LoRa, it automatically switches the
communication to GSM if the data is not updated in the database in time.

Figure 9. Dashboard showing level data received from nodes.

Similarly, actuator node sends actuator status for every ten seconds to gateway and the same will be
updated in database. If open/close command is given from webpage as shown in Figure 6, it will be
updated in the database first. Python script checks the control command from server database and
issue the open/close command to the actuator node. If actuator status is not updated in time, the
redundant GSM feature is activated. Our results show that LoRa based communication is reliable
(continuously working for 4 months), consumes low power and versatile for both indoor and
outdoor applications (gateway is installed inside the building and other nodes are installed
outdoor.).
1st International WDSA / CCWI 2018 Joint Conference, Kingston, Ontario, Canada – July 23-25, 2018

8. CONCLUSION
This paper presents a low cost IoT based solution using LoRa for monitoring and control of campus
water distribution network. Our LoRa based nodes show good coverage, energy efficiency and
reliability while reducing deployment and maintenance costs. It can be easily modified to add
pressure and flow sensor readings. GSM is used as a redundant system to ensure continuous
monitoring and control. Initial deployment results are encouraging, all remaining OHSR, UGSR
and manual valves will be instrumented to create a smart water distribution network.
Acknowledgements
This work was partially supported by the Department of Science and Technology, Govt. of India
under the Water Technology Initiative (Project. No. DST\TM\WTI\2K13\144) and the IIT Madras
Interdisciplinary Laboratory for Data Sciences (CSE\14-15\831\RFTP\BRAV).
9. REFERENCES
[1] Vavaliya Jaladhi, Bhavsar Dhruv, Kavadi Utkarsha, Mohammad Mahroof, “Online Performance
Assessment System for Urban Water Supply and Sanitation Services in India”, Aquatic Procedia,
Volume 6, August 2016, Pages 51-63.
[2] I. Stoianov, L. Nachman, S. Madden, T. Tokmouline, and M. Csail, “PIPENET: A Wireless Sensor
Network for Pipeline Monitoring,” Information Processing in Sensor Networks, 2007. IPSN 2007.
6th International Symposium on, pp. 264–273, 2007.
[3] A. J. Whittle, L. Girod, A. Preis, M. Allen, H. B. Lim, M. Iqbal, S. Srirangarajan, C. Fu, K. J. Wong,
and D. Goldsmith, “WaterWiSe@SG: A Testbed for Continuous Monitoring of the Water
Distribution System in Singapore,” in Water Distribution Systems Analysis 2010. Reston, VA:
American Society of Civil Engineers, dec 2011, pp. 1362–1378. [Online]. Available:
http://ascelibrary.org/doi/abs/10.1061/41203(425)122
[4] Z. Sun, P. Wang, M. C. Vuran, M. A. Al-Rodhaan, A. M. Al-Dhelaan, and I. F. Akyildiz, “MISE-
PIPE: Magnetic induction-based wireless sensor networks for underground pipeline monitoring,” Ad
Hoc Networks, vol. 9, no. 3, pp. 218–227, 2011.
[5] S. Jeff McKeown, “LoRaTM- a communications solution for emerging LPWAN, LPHAN and
industrial sensing & IoT applications”. http://cwbackoffice.co.uk/docs/jeff20mckeown.pdf,
[6] Libelium, “Extreme range links: LoRa 868/915mhz SX1272 LoRa module for Arduino, Raspberry
PI and Intel Galileo” http://www.cookinghacks.com/documentation/tutorials/extreme-range-lora-
sx1272-moduleshield-arduino-raspberry-pi-intel-galileo,
[7] M. Cattani, C. A. Boano, D. Steffelbauer, “Adige: an efficient smart water network based on long-
range wireless technology,” Proceedings of the 3rd .2017.
[8] Yuezhong Li, Xiaoqiang Yan, Lingyuan Zeng & Hualing Wu “Research on Water Meter Reading
System Based on LoRa Communication”, IEEE International Conference on Smart Grid and Smart
Cities, pp. 248-251, 2017.
[9] Pham, C., Rahim, A., Cousin, P.: Low-cost, long-range open IoT for smarter rural African villages.
In: IEEE International Smart Cities Conference, ISC2 2016.
https://github.com/CongducPham/LowCostLoRaGw#installing-the-latest-gateway-version
[10] Sankar, G. S., Kumar, S. M., Narasimhan, S., Narasimhan, S., Bhallamudi, S. M., “Optimal control
of water distribution networks with storage facilities”. Journal of Process Control 32, 127–137.2015.