DEVELOPMENT OF AN INTERNET OF THINGS BASED WATER MANAGEMENT SYSTEM USING DECISION TREE AND DEEP NEURAL NETWORK ALGORITHMS

ABSTRACT

Distribution of Water has been a major source of concern all over the world. Despite the fact  that  water  is  a  scarce  commodity,  a  lot  of  human  activities  in  terms  of  poor management such as opening taps when not needed and careless attitudes towards broken pipes contribute to poor distribution. Furthermore, the supply of the commodity at constant pressure to areas when not needed contributes to little or no supply to where it is needed. This is because; a lot is wasted without being used as a result of leaks and these human activities. This necessitates the need for a system to manage water distribution effectively. To this end, this research presents the Development of an Internet of Things based Water Management system using Decision Tree and Deep Neural Network algorithms. To accomplish this research work, an efficient IoT water meter was developed to take consumption  data  from  MI  Wushishi  Minna,  which  is  our  area  of  interest.  The  data generated was analyzed to understand the pattern of demand. Furthermore, a water tank capable of supplying the study area was simulated having constant valve resistance on Simulink. Based on the consumption behavior of the occupants of the estate, another simulation was done using Simulink in which the valve resistance was varied based on the demand. This results to saving water of about 3000 liters. To make the system smart, Deep neural Decision tree algorithm was used to achieve auto selection via classification. Compared to other existing work, the scheduling achieved via Decision Tree Algorithm in this research had an improved accuracy of 94.2%.

CHAPTER ONE

1.0      INTRODUCTION

1.1      Background to the Study

The importance of water cannot be overemphasized as it is widely accepted as an important necessity for life (Abu-Mahfouz  et al., 2016; Gwaivangmin,  2017; Paul, 2018). Being about 70% to 71% of the earth mass (Gupta et al., 2018; Saravanan et al., 2017), records show that 2.5% – 3.4% of it is fresh water (Rodrigues et al., 2018; Saravanan et al., 2017). Only about 0.07% – 0.08% of this is accessible for consumption (Gupta et al., 2018; Rodrigues et al., 2018). Water is used for both domestic and non- domestic activities and this is mostly drinking, irrigation and other domestic activities. For this reason, it is important to transport this essential commodity to where it is needed especially homes and industries. Hence, there is a need for the integration of so many relevant components like machines and other infrastructures, working together as a single complex and dynamic network, to aid the effective transportation of the commodity from the source to the point of use (De Corte & Sörensen, 2012; Hajebi et al., 2016). Figure 1.1 shows the water distribution network referred to as Water Distribution Network (WDN).

Several  global  efforts  have  been  made  by  various  water  utility  boards  to  ensure adequate and efficient distribution of good and quality water but it has been observed, according to Kara et al. (2016), that maintaining such a task is a global challenge. In a study by Ngancha et al. (2018), it is observed that over 348 million people experiences water scarcity. While according to the World Health Organization (WHO), 1.1billion people have no access to portable water (De Corte and Sörensen, 2016).

Ngancha et  al. (2018) in a presentation, blamed the trend on  the rapid growth of population that tripled the demand for water since 1950. This situation has been worsened  by  consistent  reduction  in  the  rate  of  rainfall  due  to  climate  change (Rodrigues et al., 2018; Turcu et al., 2012). Furthermore, other factors contributing to the  scarcity of  water  according  to  Rodrigues  et  al.  (2018),  can  be  linked  to  poor management on the part of the consumers and poor fund generation by water utility boards of various governments. It has also been reported that large portion of the fresh water distributed, flows back to the ocean unused (Saravanan et al., (2017). This is as a result of opening of taps when not needed and carefree attitudes towards leaking distribution lines. This gives a clear picture of unsustainable use of water by consumers and Water Utility Boards. This however, tends to validate the assertion by the World Bank that an estimate of 32.7 billion m3 of water is lost every year. Furthermore, it was observed that both real and apparent loss of water results to the loss of 14.6 billion USD per year (De Corte and Sörensen, 2016).

To optimize the network by minimizing losses, it is important for individuals and water utility boards to ascend a level of responsibility towards water management. This can be achieved by means of effective medium to monitor and quantify the amount of water consumed or lost (Kara et al., 2016). This however, could aid effective billing system that generates revenue and sustains the sector.

Overtime, water monitoring translating to bills has been a way to aid inclusiveness of consumers or customers and service providers in the maintenance of relevant infrastructure in order for consumers to be served efficiently. Several strategies are available for the monitoring of the usage of some basic utilities like electricity and gas with the aim of revenue generation either for expansion or maintenance of the infrastructure (Suresh et al., 2017). Most developing countries like Nigeria and Niger state in particular do not have a means of measuring the exact quantity of water consumed by users of water (Kara et al., 2016). This results to waste of resources, leading to overall hike in the cost of the process of water distribution (Rodrigues et al., 2018).

Prior to the application of meters in some nations, management of water especially in Nigeria has been done using humanitarian approach (Suresh et al., 2017). This approach characterized with a no billing system or an estimated billing system was employed since there were no appropriate measures put in place to quantify the amount of water used or lost (Abu-Mahfouz et al., 2016; Tavares et al., 2018). This may be one of the reasons why many Water Utility Boards are facing poor revenue generation (Hajebi et al., 2016) and a lot of water wastage. Furthermore, the later (estimated billing) and the recently used water meters that are either mechanical or electronic in nature still involves the frequent visit of utility officers to points of installation. This has caused frequent discord between the water service provider representative and the customers because of the issuing of estimated bills that are not justifiable. For this reason and for the sake of a better record keeping system, there is the need for a more efficient platform that aids real time monitoring of water supplied to customers without the need for new complex infrastructure leading to the transparent generation of bills. To achieve this, the Internet of Things (IoT) becomes an obvious choice (Kamienski et al., 2019).

Internet of Things (IoT) is a paradigm that allows everything and everyone to communicate through the internet (Granjal et al., 2015; Lorawan, 2017). In other words, the internet, a platform for IoT is not only expected to connect people but, also anticipated by expert to connect more than 50 billion objects by the year 2020 (Khutsoane et al., 2017). These things which may include mobile computing devices, sensors, actuators and other objects must be readable, locatable, recognizable, addressable and interconnected through the internet via a stipulated protocol called the internet protocol (IP) (Hauser et al., 2016; Patel and Patel, 2016; Li et al., 2017). This is done so as to achieve smart services and environment as shown in Figure 1.2.

This technology, enabled by other technologies such as embedded systems, wireless sensor network (WSN), radio frequency identification (RFID), cellular technology, global position remote system (GPRS) and GSM (Doni et al., 2018) is not limited by distance  and  aids  real  time  monitoring  of  water  consumed  (Lloret  et  al.,  2016). However, this could reduce frequent visits of utility representatives in the name of monitoring, aid improvement in the efficiency of water utility management, decrease operational cost, reduce customer dissatisfaction, reduce water loss (Kara et al., 2016). Furthermore,  data  generated  seamlessly could  aid  demand  forecasting  and  variable water pricing and billing (McKenna et al., 2014). In other words, unlike the traditional methods described by Bhoyar and Ingle, (2018) as labor intensive, also by Kara et al. (2016) as prone to error and characterized by discontinuous information gathering, this technology aids the continuous wireless gathering of water consumption dataset of customers in the cloud. This further enhances user experience and inclusiveness in water  management  by  enabling  each  user  the  ability  to  monitor  and  control consumption remotely at any time. This however, aids water conservation (Curry et al., 2018; Gupta et al., 2018; Lloret et al., 2016; Rodrigues et al., 2018). This means the application of this technology to metropolitan area, will aid the generation of large volume of data generated at high velocity (Suresh et al., 2017). In the long term, this data collection could foster demand prediction, detection of incidence such as tampering and leak, characterization of customers (Lloret et al., 2016) and automatic execution of decision to curb damages, waste and ill services (Ahmed et al., 2017).

In summary, the need for real time dataset of water consumed is necessary for effective billing and for smart techniques used for network optimization. This however, leads to revenue generation that aids sustainability of water infrastructure. This can be efficiently achieved via the use of IoT as a platform to monitor consumption and to aid consumer inclusiveness in water management. Furthermore, this data will aid analysis and optimization techniques that will result to better water management and services. However, despite the positive effect of this technology, its application may be rather challenging in developing nations like Nigeria where some part of the country is characterized with erratic power supply. This will however, increase the probability of outage resulting in large loss of data aside the few times when downtime is experienced within the wireless network. To solve this problem, this study presents the design and development of IoT based water monitoring system powered by a Green source.

In a WDN, effective distribution is achieved via the use of two main approaches. The first approach  involves  the  use  of  electric  pump  (Balekelayi  &  Tesfamariam,  2017)  which pumps water from a reservoir of treated water usually underground, to the consumers. This however, seems inefficient because of the high power consumed for continues operation (Abdallah, 2020). According to Tsai et al. (2018), 30% to 40% of power of a metropolitan city is used to pump water. From this consumption, 80% to 90% of the power is absorbed by motor pump set (Sarbu, 2016). This however, increases the cost of operation (Sarbu, 2016). To further reduce the cost of operation, researchers have proposed and employed the second approach which is the use of elevated storage tanks mounted on calculated heights as shown in Plate I so as to achieve effective distribution, leveraging on pressure due to gravity (Muhammad & Safdar, 2020; Ree & Eddy, 2016; Torkomany & Abdelrazek, 2020). Usually, treated water served, had to be pumped to the tank for storage before being distributed.  With  this,  required  pressure  is achieved via  gravity to  serve  consumers if properly designed. This however, eliminates the costly frequent pumping operation. However, even with this, the aging infrastructure most times, impedes optimal use of the network since the aged network is characterized with leaking pipes (Balekelayi & Tesfamariam, 2017; Brooks et al., 2018). Most times, as a result of the inability to predict failure, reactive maintenance is often times given to the network. This however, does not ensure efficiency in distribution as many leaks can be noticed at the same time by different people in different location. This leads to low pressure and waste (Balekelayi and Tesfamariam, 2017) that will increase the frequency of pump operation if the whole network is characterized with constant maximum pressure.

Logically, a total overhaul of the network may be suitable to cure the ailing network. However, the expenses to be incurred will be overwhelming. In the United States alone, $3.6 trillion dollar will be needed by 2020 to fix the network (Brooks et al., 2018). This may not be feasible in some developing nations as the annual budget may not be enough to undergo such project. This therefore, suggests that there is a need to optimize the design from the beginning of the development of a network and device a means to optimize the operation of the network to achieve acceptable pressure based on demand to aid water conservation on the long run.

Over the years, scheduled pumping operation in WDN has been used to aid water conservation and to reduce the cost of energy used in pump operation (Brentana et al., 2017; Sarbu, 2016; Stokes et al., 2016). This helps utility not to run into water shortage (Telles et al., 2016). Generally, the operation is done via the use of district state pumping schedule. In the approach, the pump is turned on when needed and turned off when not needed (Abdallah, 2019; Archetti et al., 2018). However, this may not be healthy for distribution since the demand curve in the network is not discrete. In other words, there can’t be a time when no water is needed in the network. Furthermore, for distribution method that employ the use of elevated tanks, maximum pressure is delivered as a result of gravity to all part of the network even if not needed. This causes waste in areas where fault is discovered. Furthermore, as the water level reduces, the extreme part of the network may be deprived of good quantity of water when needed since every part of the network is opened at the same pressure rate to the flow. However, for effective scheduling, the consumption behavior of the consumers must be considered (Torkomany &  Abdelrazek, 2020) in proffering solutions. This can be easily achieved via classification of valve resistance based on consumption pattern in the data collected by metering devices. By observation, it is noticed that in societies like Nigeria two major classes of customers exist. This includes the residential and nonresidential customers which includes offices, marketplace and schools. Usually, it is observed that there is population movement from points of residence to points of non-residence and back. This however, increases the dynamism of the demand curve in both residential and nonresidential areas. In other words, there are times of the day that more population is seen at residential areas while the nonresidential areas have less population and vise vassal. As a result, a behavioral pattern can be observed and used to channel more water to where is most needed and less water where it is least needed. This will be effective via the use of transparent mediums in collecting data such as IoT based water meters. The understanding and use of these data will ensure that water is not totally cut off when demand is low at some point. To this end, this research presents development of an Internet of Things based water management system using decision tree and deep neural network algorithms.

1.2 Statement of the Research Problem

In Nigeria, the use of elevated tank system of distribution is ideal because of the inability of the ailing power sector to support continuous pumping of water in the WDN. As a result of the growing population, the need for more tanks is inevitable. However, the adaptation of old designs involving constant peak pressure discharge through a channel to the whole network may not give optimal performance of the network. This is because, as noted, the whole network could be classified as residential, nonresidential or both and therefore is characterized by varying volume of population at different points in time characterizing the network with varying demand at all times (Torkomany & Abdelrazek, 2020). For example, water demand in residences is more during the morning and evening hour (Sarbu, 2016). Therefore, pumping at constant high pressure (Sarbu, 2016), may not be sustainable if leaks are noticed in-between. Furthermore, huge losses may occur as a result of opened taps and broken lines in the distribution when water is supplied at constant high pressure to areas of low demand. This however may result to customer dissatisfaction and losses that increases the frequency of pumping, making the operation of a WDN costly. To cushion this ill, this research presents the Development of an Internet of Things  based Water Management System using Decision Tree and Deep Neural Network Algorithms. In this presentation, IoT based monitoring device will be used to study the latent behavioral pattern as relate to consumption. Furthermore, various classes of valve resistance are predicted for automatic operations using the decision tree algorithm so as to optimize the distribution system.

1.3 Aim and Objectives of the Study

The aim of this research is to develop an Internet of Things based water management system using decision tree and deep neural network algorithms. This was achieved by the following to:

i.      Develop an efficient IoT based monitoring device (water meter) to study the consumption pattern in the location where it is installed.

ii.     Simulate the daily pumping of water at constant valve resistance on Simulink using the leaking tank model.

iii.      Simulate water pump operation with varying valve resistance to supply water based on demand using Simulink.

iv.      Develop a smart distribution network using decision tree and deep learning algorithms.

v.     Comparative analysis and performance evaluation of objective ii, iii and iv.

1.4      Justification of the Research

In several parts of the world today, the main objective of utility is to ensure efficient water distribution (Abdallah, 2020). One of the factors to measure efficiency remains the resilience of water delivery (Sarbu, 2016). However, one way to achieve this is the optimization of pump operation (Abdallah, 2020). This has been looked into by several researchers such as Torkomany and Abdelrazek (2020), Candelieri et al. (2020) and Abdallah (2019). The submission narrates how different techniques have been used to optimize water supply. One of such includes pump scheduling. However, based on review so far, the approach of variable valve resistance using the demand behavior of consumers in a location to fairly schedule how much water is to be supplied as used in this research, have not been done. This therefore, justifies this research.

1.5      Scope of the Study

The  scope  of  the  research  work  is  restricted  towards  water  conservation  and consumption accountability achieved  via the development  of  IoT  water monitoring device and smart pump operation optimization that aids reliance in WDN in terms of water delivery in M.I Wushishi estate. The node was connected to two different houses on two different working days in the week to understand the pattern of water consumption in those houses. House ‘A’ contains four adults employed with the state government, one house maid who is a teenager and three children in primary school. House ‘B’ consists of just two adults. The proposed standards made by WHO of water required by individuals was then employed as the bases for simulation in Mat-lab. This helped to generate data used for pump operation optimization, resulting to varying flow resistance as a result of varying valve resistance.

₦8,000.00 – DOWNLOAD FULL PROJECT DOWNLOAD FULL PROJECT Added to cart

---

---

---

---

Related Project Topics :