South Asian Journal of Engineering and Technology Vol.3, No.

4 (2017) 105–117

ISSN No: 2454-9614

Monitoring and Fault identification of three phase Transmission line

using Internet of Things (IOT)
L.Sakthivel, V.Logash, G.Tamil selven, C.Mohandhass
GNANAMANI COLLEGE OF TECHNOLOGY, Tamilnadu, India.

Received: 12/01/2017, Revised: 16/02/2017 and Accepted: 27/03/2017

Abstract

Fault detection and location has been a goal of power system engineers since the creation of distribution and
transmission systems. Quick fault detection can help protect equipment by allowing the disconnection of faulted lines before any
significant damage is done. Accurate fault location can help utility personnel remove persistent faults and locate areas wher e
faults regularly occur, thus reducing the frequency and length of power outages. As a result, while fault detection and location
schemes have been developed in the past, a variety of algorithms continue to be developed to perform this task more accuratel y
and more effectively. The detection and location of faults on power transmission lines is essential to the protection and
maintenance of a power system. Most methods of fault detection and location rely on measurements of electrical quantities
provided by current and voltage transformers. These transformers can be expensive and require physical contact with the
monitored high voltage equipment.

1. Introduction
Many electric power transmission companies s have primarily relied on circuit indicators to detect faulty
sections of their transmission lines. However there are still challenges in detecting the exact location of these faults.
Although fault indicator technology has provided a reliable means to locate permanent faults, the technical crew and
patrol teams still has to physically patrol and inspect the devices for longer hours to detect faulty sections of their
transmission lines.Wireless sensor based monitoring of transmission lines provides a solution for several of these
concerns like real-time structural awareness, faster fault localization, accurate fault diagnosis by identification and
differentiation of electrical faults from the mechanical faults, cost reduction due to condition based maintenance
rather than periodic maintenance, etc. The use of sensor networks has been proposed for several applications like
mechanical state.These applications specify stringent requirements such as fast delivery of enormous amount of
highly reliable data. The success of these applications depends on the design of a cost-effective and reliable network
architecture with a fast response time. The network must be able to transport sensitive data such as current state of
the transmission line and control information to and from the transmission grid. This research provides a cost
optimized framework to design a real-time data transmission network. Our objective is to formulate a

105
 South Asian Journal of Engineering and Technology Vol.3, No.4 (2017) 105–117

communication frame-work to transport enormous amount of sensitive data at the time scale of Supervisory Control
and Data Acquisition (SCADA) cycle.
To monitor the status of the power system in real-time, sensors are put in various components in the power
network [7]–[9]. These sensors are capable of taking fine-grained measurements of a variety of physical or electrical
parameters and generate a lot of information. Delivering this information to the control center in a cost efficient and
timely manner is a critical challenge to be addressed in order to build an intelligent smart grid [2]. Network design is
a critical aspect of sensor based trans-mission line monitoring due to the large scale, vast terrain, uncommon
topology, and critical timing requirements. The goal is to deploy multiple different sensors in critical locations of the
transmission line to sense mechanical properties of its various components and transmit the sensed data through a
suitable wireless network to the control center. At the control center, it can be combined with existing electrical data
in the system to arrive at an ideal preventive or corrective control decision. We design a hybrid hierarchical network
that spans wired, wireless and cellular technologies to provide cost optimized delay and bandwidth network.
This paper is organized as follows. Section II explains the related work followed by sensor network design
in Section III. Section IV presents evaluation studies and Section V concludes the paper.

Block Diagram
Transmission line infrastructure, wireless networking presents a feasible and cost effective solution for
transmission of monitoring data [3]. Several works [11]– [15] and [16] propose to improve the state of the art in
transmission line monitoring by harnessing the power of wireless sensor networks for real-time monitoring and
control.

The linear network topology proves to be a major challenge for wireless network design with respect to
latency constraints and bandwidth constraints. Performance evaluation of the linear network model [20] shows that
successful delivery ratio of the packets from the nodes far away from the substation is found to be much less than

106
 South Asian Journal of Engineering and Technology Vol.3, No.4 (2017) 105–117

that of nodes near the substation because packets from a farther node have to travel a longer distance and the rate of
collision is higher. The effective monitoring of a large transmission line network requires a hybrid communication
infrastructure.

This hybrid infrastructure can be a combination of wired (copper cable/optical fiber) and wireless
(cellular/IEEE 802.15.4) standards to enhance the capability of the overall network to meet newer requirements
based on emerging smart grid applications.

In this paper, we formulate a hybrid hierarchical network de-sign problem that can provide cost effective
data transmission while at the same time respecting the bandwidth, delay, and connectivity constraints. We
formulate a placement problem to optimize the number and location of the cellular enabled towers to significantly
reduce the operational and installation costs while respecting all the constraints.

Voltage Sensing

Power Supply

A. Three Level Hierarchical Network

107
 South Asian Journal of Engineering and Technology Vol.3, No.4 (2017) 105–117

We propose a hierarchical three level wireless network model for time critical applications. Each level is
equipped with an array of sensors and transceivers with varied capabilities such that together they achieve the
required behavior. The design involves the installation of a private WSN of low cost, low data rate links, utilization
of the existing SCADA network, and a wide area network such as cellular network comprised of ex-pensive but high
data rate links. The proposed network makes use of the existing SCADA links (optical fiber) for communication
between substations and control center and strategically utilizes the existing cellular network for data transmission
from certain transmission towers directly to the control center. A set of wireless sensors on each tower is installed as
part of the private WSN.
Fig. 2 depicts a power transmission corridor with a number of transmission towers, two substations, one at
each end of the transmission line, and a control center.

EMBEDDED SYSTEM
An embedded system is a combination of computer hardware, software and additional mechanical parts,
designed to perform a specific function .An embedded system is designed to do a specific task within a given time
frame, repeatedly, without human interaction. Embedded system do not need a complete operating system, but only
the basic functionalities of an operating system in a real-time environment, that is, a real time operating system.
(RTOS). Frequently, embedded system does not have a user interface.

PIC (PERIPHERAL INTERFACE CONTROLLER)

PIC(Peripheral Interface Controller) is the IC which was developed to control the peripheral device,
dispersing the function of the main CPU.When comparing to the human being, the brain is the main CPU and the
PIC shares the part of which is quivalent to the automatic. However, the through out, the memory capacity are not
big. It depends on the kind of PIC but the maximum operation clock frequency is about 29 MHZ and the memory
capacity to write the program is about 1k to 4k words. It is possible to make the compact circuit when using
PIC.forms a cluster supporting many to one communication from all the nodes in the cluster to the cluster head.

The first level of the network is responsible for collecting in-formation about the tower. It is composed of
sensor nodes in-stalled in each transmission structure forming a sensor array in tower (SAT). This SAT consists of
an array of sensor modules such as tension sensors, accelerometers, temperature sensors, tilt sensors, motion
sensors, vision-based sensors, and infrared sensors, etc., similar to [13]. Each tower is equipped with a more
sophisticated relay node with enhanced computation and communication capabilities. Data from each sensor in the
SAT is transmitted to the relay node. The relay node is responsible for compressing the data received from the SAT
and transmitting it to the higher level.
The second level of the network is responsible for transmission of data from towers that are far away from
the substations. Consider a segment composed of a few towers in the middle of the transmission line network. Data

108
 South Asian Journal of Engineering and Technology Vol.3, No.4 (2017) 105–117

from these towers cannot reach either of the substations due to limited bandwidth of the intermediate wireless links.
In such cases, enabling one of these towers with cellular capability can provide a feasible solution as shown in Fig.
2. It is to be noted that it is not required to enable all towers with cellular technology as proposed in [11]. The
second level is thus composed of segments of such towers transmitting their aggregated information to the cellular
enabled transmission tower which acts as the head of their segment. The cellular enabled tower is a transmission
tower equipped with an additional cellular transceiver along with the relay node. This cellular transceiver offers an
alternative way to deliver the tower’s data directly to the control center through a high band-width, low latency
cellular network.
The Coverage Cons trained placement problem formulation requires little modification from our original
placement problem formulation. Let denote the set of nodes not covered by the cellular service. The constrained
edge set, will be. In order to model the Coverage Con-strained placement problem, the input graph needs to be re-
placed by . Fig. 4 shows the example where towers 3 and 4 do not have any cellular coverage. The edges (3, CC)
and (4, CC) are thus removed. In the formulation, the corresponding binary variables, and are removed to represent
cellular unavailability at these towers. Thus, with a simple modification in the formulation, restricted cellular
availability can be easily addressed.
If there is a large segment of towers devoid of cellular coverage, then the responsibility of data delivery
falls upon wireless Zigbee links. In such cases, the data delivery might not be able to meet the latency constraints
given the limited capacity of wire-less Zigbee links. Such cases can be easily identified though our proposed method
as shown in Fig. 10.

Fig. 5. Accommodating asymmetric flow requirements

F. Asymmetric Data Generation

The method proposed in [12] relies heavily on symmetry and hence cannot accommodate asymmetric flow
bandwidth requirement in the network. There can be several scenarios leading to towers generating sensor data at
different rates. This can be due to a requirement of fine grained sensor measurement in order to attain better
situational awareness of a particular tower located in a sensitive area. Fig. 5 shows such a scenario. Our proposed

109
 South Asian Journal of Engineering and Technology Vol.3, No.4 (2017) 105–117

formulation can easily accommodate such asymmetric requirements. This is because each flow generated at a tower
is individually formulated. Asymmetric data generation will change the values of and in (2) and (7). Hence the
formulation solves both symmetric and asymmetric cases equally the only difference being in the input file
specifying the flow requirements.

G. Incremental Deployment

With changing data traffic requirements and geographical expansion of the transmission line, the possibility
of incremental deployments is always present. If at each such future requirement, an entirely new input is given to
the optimization formulation without taking into account the currently existing cellular enabled towers, the solution
may be a costlier deployment. We term this method as memoryless deployment because it discards any memory of
existing cellular enabled towers. This method can result in installation on an entirely new set of towers, thus risking
loss of any investment made in installing existing cel-lular transceivers in the first place.

We present a method to add new cellular links on top of existing network to satisfy newer requirements
while minimizing installation costs. In order to do so, only the input to the optimization program needs to be
modified. The installation costs for currently deployed towers is made equal to 0 and flow requirements are
increased as per new specifications leading to modified latency calculations. Since the installation costs of cur-rently
deployed towers is now 0, they will be automatically considered as cellular enabled towers in the optimized answer.
The optimization formulation will make as much use of such existing cellular enabled towers as possible in
minimizing total costs. Thus, incremental deployment reuses the existing cellular enabled transceivers as much as
possible.

4. Performance evaluation
We consider a transmission line network with 75 towers with an average span length of 800 ft [23]. In
order to reflect real-world scenarios, the bandwidth of the optical fiber links (SS, CC) is taken as 10Gbps.
Bandwidth of the cellular links is taken as 75 Mbps and latency incurred in the cellular link due to state transition
delay, access delay, and handover, etc., is taken as 50 ms [25]. The bandwidth of the IEEE 802.15.4 wireless links is
250 kbps and latency incurred due to these links is 16 ms [24]. The length of the data packet generated by each
tower is 32 kbits [11]. The performance metric of interest is the total cost of the network including the installation
and operational costs. We consider three pricing schemes named and analyze the effect of different costs associated
with each type of link present in our network. They vary in their ratio of operational costs attached to each type of
link. A pricing scheme can be described as a ratio of operational costs of optical fiber to cellular to Zigbee. Thus, a
scheme (1:10:2) would mean that operational cost of the three types of links: optical fiber, cellular and Zigbee are in
the ratio 1:10:2.

110
 South Asian Journal of Engineering and Technology Vol.3, No.4 (2017) 105–117

We study several different scenarios including variation in flow bandwidth, end to end deadline and
network size. We compare the results of our proposed formulation called Integer Linear Program (ILP) with the
method proposed in [12] referred to here as the Quadratic Equation method (QE). We also evaluate the cost of the
network in cases of constrained cellular coverage and incremental deployment.
The proposed formulation provides an optimal solution to the wireless network design problem at the
granularity of a trans-mission corridor as shown in Fig. 2. This solution can be in-dependently computed for each
corridor. While computation-ally expensive, these calculations are required to be done only during the offline
centralized network planning stage. Thus, the fact that an ILP formulation is expensive to compute is mitigated by
the small number of times such an optimization needs to be performed. However, for cases where the transmission
corridor comprises of several hundreds of transmission towers, heuristic methods may be better suited. This is
because even though they compromise on the accuracy of the result, they are capable of providing a solution with
much less computing re-source requirements. Although the ILP model is harder to scale and computationally-
intensive, it is helpful in determining a lower bound on the costs. We use the ILOG CPLEX 12.2 soft-ware [21] to
solve the proposed Integer Linear Program (ILP).

Fig. 6. ILP: Effect of variation in flow bandwidth.

The simulations were run on Intel(R) Xeon(R) X5650@ 2.67 GHz machines. The simulations took a
minimum of 0.16 seconds and a maximum time of 4.25 hours.

A. Efect of Variation in Flow Bandwidth

Fig. 6 shows the effect of the amount of data generated by each tower and its effect on the feasible
operation of the trans-mission line. In this simulation, we consider a 75 node network with a deadline constraint of 3

111
 South Asian Journal of Engineering and Technology Vol.3, No.4 (2017) 105–117

s. The packet size of the sensor data generated at each tower is the same. Given constant band-width of the wireless
links, only a certain number of flows can be multiplexed on any link. Thus smaller the flow bandwidth requirement,
more flows can be multiplexed on each link. Given a large deadline requirement, this results in reducing the number
of cellular links to be used and hence the cost. The performance graph echoes this observation. Notice in the graph
that for values of flow bandwidth greater than or equal to 128 kbps (specifically 128 kbps, 160 kbps, and 192 kbps),
the cost becomes constant. This is because given the wireless link bandwidth of 250 kbps, at most one flow can be
multiplexed on each link. Thus the network design remains same for each of these three flow bandwidth
requirements. Also in cases where, flow bandwidth is greater than wireless link bandwidth, then the remaining
options are either deploying an all cellular or all wired solution.

Fig. 7 compares the results given by our proposed algorithm, ILP and the proposed method (QE) in [12]
with respect to variation in flow bandwidth. As mentioned earlier, the QE method [12] utilizes a quadratic equation
to obtain the number of cellular enabled towers. Roots of quadratic equation are rounded off to the nearest integer to
depict the number of cellular enabled towers which must be an integer. This rounding off leads to incorrect results as
can be seen in the plotted curves. Once flow bandwidth is greater than 128 kbps, the results should be the same for
the three cases (128 kbps, 160 kbps, and 192 kbps) as explained previously. However, the QE method tends to give
incorrect results owing to errors encountered due to rounding of roots. Similarly for flow bandwidths of 84kbps and
96kbps, the cost incurred by QE method is less, but that is because number of towers selected by the QE method are
insufficient leading to constraint violation.

112
 South Asian Journal of Engineering and Technology Vol.3, No.4 (2017) 105–117

E. Efect of Operational Period

Fig. 11 shows the effect of operational period on the total costs (initial installation costs and operational
costs over the operational period). In case of fixed operational costs, total costs increase rapidly with increasing
operational period. In case of adaptive costs depending upon link utilization the total costs re-duce dramatically.
Also in this case, if the available link band-width is much higher as compared to bandwidth utilization, then
operational costs turn out be very less and over time the initial investment on installation is recovered. Thus it can be
profitable to deploy a cellular transceiver on a bigger subset of towers. However, if cellular bandwidth is relatively
less, making the link utilization based operational costs a bigger part of the total costs, then the number of cellular
towers needs to be optimized. This graph illustrates how different factors such as available link bandwidth,
operational period and cost ratios affect minimum cost network design.

F. Effect of Link Utilization Based Cost

Fig. 12 shows the impact of link utilization based adaptive cost models on total network design cost. In this
simulation, we consider a 75 node network with an operational period of 100. We consider 4 cost models explained
as follows.

The Fixed Cost model considers fixed subscription cost ignoring adaptive link utilization. As evident, it
incurs highest costs because the number of deployments increase within-creasing flow bandwidth requirement.
The Adaptive Cost, Link BW(75 Mbps) model is the link utilization based cost model where cellular
bandwidth is 75 Mbps incurring the lowest costs. The optimal solution here is to deploy all towers with cellular
capability. This solution is optimal due to two factors. One, the subscription costs become negligible as compared to
wireless operation costs given the high cellular bandwidth availability as compared to utilization. Secondly, a huge

113
 South Asian Journal of Engineering and Technology Vol.3, No.4 (2017) 105–117

initial investment can be easily recovered given a large operational period. The negligible subscription cost acts as
an incentive to initially deploy all towers with cellular capability.

In order to consider the effect of available cellular bandwidth, we consider an Adaptive Cost, Link BW(0.5
Mbps) model with a smaller available link bandwidth of 0.5 Mbps. In this case link utilization based subscription
costs are non negligible resulting in increased cost.

We consider another cost model, Adaptive Cost, Base Subs where in addition to the link utilization based
adaptive costs, a fixed base subs cription fee is charged per active cellular link per period. The available cellular
bandwidth is 75 Mbps. Thus subscription costs are negligible but base subscription fee acts as a major factor in cost
calculations. When compared with Adap-tive Cost, Link BW(0.5 Mbps), the tradeoff can be observed. At lower
flow bandwidth requirements,

Adaptive Cost, Base Subs incurs more cost due to overpowering base subscription fee. But at higher
requirement, link utilization based costs result in more costs by Adaptive Cost, Link BW(0.5 Mbps).

G. Efect of Link Unreliability

Fig. 13 shows the impact of link unreliability on network de-sign cost. In this simulation, we consider a
constant flow band-width of 32 kbps and a deadline constraint of 3 s. At lower link reliabilities, a path should
consist of lesser number of links to maintain path reliability constraint. This leads to more cellular towers being
deployed resulting in higher costs. As link reliability increases, more wireless links can be utilized resulting in cost
reduction. After a certain point (maximum links ), any further increase in link reliability does not affect cost
reduction. This is because

114
 South Asian Journal of Engineering and Technology Vol.3, No.4 (2017) 105–117

each tower generating sensor data at the rate of 32 kbps initially. The data generation rates are then
gradually increased upto 128 kbps to mirror the increasing bandwidth demands in the future. Fig. 14 shows the
cumulative costs incurred by the two methods: incremental deployment and memoryless deployment when the
operational period is equal to one. Memoryless deployment starts with a clean slate each time a new requirement
comes in. Due to this, memoryless deployment ends up in-stalling a cellular transceiver on a much bigger subset of
towers.

Please note that both deployment methods are variations of our proposed formulation producing optimal
solutions de-pending on the input being provided to them. Our formulation offers the flexibility to perform different
types of analyses. Factors such as operational period and link utilization based costs affect the performance of the
two methods. Consider Fig. 15, where operational period is varied from 20 to 350. We consider a cost model
[C1(1:10:2)] with the variation that each active cellular link is charged a base subscription fee and link utilization
based subscription cost instead of fixed subscription cost. Each point on the graph, represents the total costs of an
instance where a set of data requirements (starting from 32 kbps to 128 kbps) was operational for a period . Each
value on the curve is the total cost of memoryless deployment normalized with respect to the respective value for
incremental deployment to show their relatively close but different performance.

In Fig. 15, we observed that at low operational periods, memoryless deployment incurs more cost than
incremental deployment. This is because memoryless deployment ignores any in-vestment made in the earlier
deployment and a small operational period does not allow enough time to recover from that investment. Incremental
deployment on the other hand, avoids extra installations by making the most use of the already deployed towers. At
higher operational periods, the Fig. 15. ILP: Comparing Memoryless deployment with Incrementa deployment.be
recovered and hence memoryless and incremental deployments produce solutions with equal costs.

115
 South Asian Journal of Engineering and Technology Vol.3, No.4 (2017) 105–117

4. Modeling Multiple Constraints

In the proposed formulation, we presented a generalized formulation in order to encompass varied
application scenarios. However, it can be difficult to model every operational constraint. Mainly, the most important
constraints that need to be modeled for a feasible network design are flow latency and flow bandwidth constraints.
This is because Zigbee links can provide the least expensive communication, but their limited bandwidth proves to
be a major bottleneck. Also the topology of trans-mission lines presents critical challenges towards low latency
communication being achieved solely by Zigbee links. These are followed by the cellular coverage constraint since
extreme unavailability of cellular coverage might render the network de-sign infeasible.
Other constraints such as link reliability, link utilization based cost structure, asymmetric data generation,
etc., are less critical for a feasible cost optimized network design. The results we present in the paper individually
analyze the effect of variation in each of the constraint on the cost of the network while other constraint are assumed
to take up average values. Specifically, we find that network design is feasible at very low costs at higher latency;
and at very high costs at higher flow bandwidth. Further, we find that if link utilization based cost structure is
adopted then flow bandwidth requirement might have no effect on the total costs as shown in Fig. 12.

5. CONCLUSION

In this paper, we presented an optimal formulation for a cost optimized wireless network capable of
transmission of time sensitive sensor data through the transmission line network in the presence of delay and
bandwidth constraints. Our analysis shows that a transmission line monitoring framework using WSN is indeed
feasible using available technologies. We compare behavior of our proposed method with the method proposed in
[12]. Our proposed formulation is generic and en-compasses variation in several factors such as asymmetric data
generation at towers, wireless link reliabilities, link utilization dependent costs, non-uniform cellular coverage
characteristics and requirements for cost optimized incremental deployment. Our evaluation studies show that the
main bottleneck in cost minimization is wireless link bandwidth. Further, in cases of increasing flow bandwidth, the
limited wireless link bandwidth leads to a feasible but expensive design due to increased dependence on cellular
network to satisfy constraints. As part of future work, we plan to study cost efficient fault tolerant network design.

References
[1] H. Li, G. W. Rosenwald, J. Jung, and C. Liu,
“Strategic power infra-structure defense,” Proc. IEEE, vol. 93, no. 5, pp. 918–933, May 2005.
[2] P. Zhang, F. Li, and N. Bhatt, “Next-generation monitoring, analysis, and control for the future smart control center,” IEEE Trans.
Smart Grid, vol. 1, no. 2, pp. 186–192, Sep. 2010.
[3] V. C. Gungor and F. C. Lambert, “A survey on communication net-works for electric system automation,” Comput. Netw., vol. 50, no.
7, pp. 877–897, May 2006.
[4] P. Ramachandran, V. Vittal, and G. T. Heydt,

116
 South Asian Journal of Engineering and Technology Vol.3, No.4 (2017) 105–117

“Mechanical state estimation for overhead transmission lines with level spans,” IEEE Trans. Power Syst., vol. 23, no. 3, pp. 908–915,
Aug. 2008.
[5] S. Malhara and V. Vittal, “Mechanical state estimation of overhead transmission lines using tilt sensors,” IEEE Trans. Power Syst., vol.
25, no. 3, pp. 1282–1290, Aug. 2010.
[6] J. Ausen, B. F. Fitzgerald, E. A. Gust, D. C.
Lawry, J. P. Lazar, and R. L. Oye, “Dynamic thermal rating system relieves transmission con-straint,” in Proc. IEEE 11th Int. Conf.
Transm. Distrib. Construction, Oper., Live-Line Maintenance (ESMO), Oct. 15–19, 2006.
[7] V.C.Gungor,L.Bin,andG.P.Hancke,“Opportunities and challenges of wireless sensor networks in smart grid,” IEEE Trans. Ind. Electron., vol.
57, no.10,pp.3557–3564,Oct.2010.

117