You are on page 1of 6

PERFORMANCE EVALUATION OF CURRENT

DIFFERENTIAL RELAYS OVER A WIDE AREA NETWORK


P. Beaumont*, F. Kawano*, A. Kawarada, T. Kase, H. Sugiura
F. Lam, J. Hurd, P. Worthington, D. Richards, P. Merriman
*Toshiba International (Europe) Ltd, U.K.

Toshiba Corporation, Japan

Alcatel-Lucent, Canada
philb@til.toshiba-global.com

Keywords: Ethernet-based-protection, Communications


infrastructure,
Next-Generation-Networks,
Wide-areaEthernet.

Abstract
Progress in the development of communication network
technology continues apace and Ethernet technology, once
limited to local area network applications, is increasingly
being considered for use as the backbone technology for wide
area communication networks. The technology, most
commonly referred to as Carrier Ethernet, seems likely to
become the dominant technology within wide area networks
ultimately replacing earlier generation networks based upon
PDH and SDH / SONET platforms. The availability of
Carrier Ethernet and in particular, in terms of the testing
described here, IP/MPLS-based Carrier Ethernet now enables
us to consider the application of Ethernet technology to
teleprotection.

1 Introduction
Since 2007 we have continued to make progress with the
development, manufacture and testing of a gigabit Ethernet
based current differential relay [1, 2]. Although confident that
our internal evaluation had been thorough we were cognizant
of the fact that the communications infrastructure used for our
internal evaluation is a local network facility. Accordingly,
we recently undertook a joint evaluation with Alcatel-Lucent,
in which we were able to prove that a gigabit Ethernet
protection relay system operating over a wide-area network
has the capability to be utilized in a practical, real world
environment.
Teleprotection is an essential technology for operating and
maintaining a reliable, robust and safe electric grid.
Teleprotection devices rely upon a deterministic service
provided by a stable, symmetric, constant delay
telecommunications network for their communication
requirements. Current differential protection, widely applied
for the protection of HV and EHV feeders due to its inherent
strengths of high sensitivity and selectivity, relies upon the
provision of a relatively high-bandwidth communication
channel and the existing PDH and SDH / SONET networks
have proven to be well suited to this task. In view of the more

demanding performance imposed upon the communications


network, current differential protection was chosen as the
primary teleprotection to lead this investigation.
In order to assess the suitability of Ethernet technology for
teleprotection the authors established a demonstration system
to study the operation of current differential relays over
Ethernet. A solution to the fundamental requirement of
establishing and maintaining synchronisation between relays
was developed using demonstration relays incorporating a
newly developed, dedicated Ethernet interface with an
integrated high-accuracy time control function. Based upon a
comprehensive set of tests [3], it was concluded that a
network comprised of IP/MPLS routers will comply with all
of the requirements of the demonstration relays.

2 Current Differential Protection over Ethernet


A current differential protection relay compares locally
measured current data with data from the remote end of the
transmission line which has been transmitted from the remote
terminal via a communication channel. It determines that an
internal fault condition has occurred when a differential
current results from the comparison of the two current data
quantities. In order to compare two current data quantities
correctly, the data must be representative of the same instant
in time. Since a relay cannot compare the current values until
the remote end data arrives, the transmission delay must be
short in order to achieve fast operation of the protection. Data
communication is vital for the operation of current differential
protection, and the requirement on performance of the
communication channel is correspondingly high.
The requirements of the current differential relay on the
communication channel are two-fold as follows:
-

The required operating time of the protection relay


must be met
Synchronisation of the data must be achievable

In order to achieve the required protection relay operating


time, the transmission delay time through the communication
channel is critical, and generally the required operating time
is shorter at higher system voltage levels.
Each relay requires sampling timing synchronisation. In order
to be able to achieve synchronisation using data

communication, communication performance such as


throughput and jitter significantly affects the accuracy of
sampling timing synchronisation. Dedicated communication
systems with limited bandwidth have been applied to existing
current differential relays. Thus, protection systems have been
realized by elaborating methods such as time compression
multiplexing transmission.
In recent years power utilities have been investigating the
possibility of applying common communication facilities
across a broad number of services. It is highly likely that
Ethernet based communication, superior in terms of its highspeed performance, flexibility and maintainability, will be
applied for protection. By making the best use of networks,
we will be able to make a significant contribution to the
realisation of next generation protection relay systems that are
able to adopt unified telecommunication services for
protection relaying. Benefits include the ability to configure
cost-effective means of redundancy, simplification and
reduction of site work, and the practical application of novel
protection algorithms that utilize the increased amount of
information available. Toshiba has undertaken an evaluation
on trial current differential relays which can be applied to
Ethernet [1]. We envisage that the application of wide-area
networks to communication for protection enables multiple
devices on the network to share large amounts of information
at higher speed and to contribute to enhancement of the
performance and extension of functions.
However our evaluation utilizing trial products had been
exercised within an ideal communications environment
without any disturbances. An evaluation within a wide-area
network environment, as experienced in the area of control
systems, is a key requirement in achieving practical
application of the technology. This paper introduces the
evaluation results using our trial products connected to an
IP/MPLS (Multi-Protocol Label Switching) network as
applied to wide-area networks.

performance of teleprotection devices. Whereas this notion


may apply to links between devices that are carried over the
Public Internet and thereby compete with all kinds of traffic
for network resources, this notion is not applicable when the
Utility operates a private network based upon IP/MPLS
and, as with a traditional PDH/SONET/SDH network,
suitable care is taken in the engineering and provision of the
network.
3.1 Areas for evaluation
The areas to be evaluated were selected so that the effects
generated by the connection configuration of the devices and
their impact or disturbance on the line differential relays
could be evaluated. Of special interest was behaviour related
to transmission delay, jitter, transmission errors, and network
reconfiguration:
-

Number of routers in the path


Transmission delay and jitter
Transmission burden (disturbances, bandwidth
restrictions)
Route switching in response to communication
failures

3.2 Test configuration


The test arrangement is shown in Fig. 1. The different
equipment types, and their purpose in these tests, are briefly
described as follows. The IP/MPLS communication network
is composed of several IP/MPLS routers (manufactured by
Alcatel-Lucent, hereinafter referred to as MPLS routers) and
1Gbps communication circuits.

3 Testing
Legacy networking utilized time-division multiplexed
transmission architectures based on PDH/SONET/SDH to
provide the communication channel between units. The
circuit-switched nature provided some guarantee of delay
limit, delay stability, and transmission symmetry. In order to
support next-generation interfaces and applications, the
network architecture has to change in order to support the
packets/frames associated with these new interfaces. From a
telecommunications network perspective it is necessary to
provide the connectivity between devices with a high quality
of service quantified by limits on transmission delay, jitter,
and asymmetry. These are, obviously, in addition to
reliability, robustness, manageability, and ease of use.
Next Generation Networks (NGNs) are based upon packetswitching architectures. One of the perceived drawbacks of
NGNs, as applied to teleprotection schemes, is the notion that
the statistical nature of packet networks will impact upon the

Figure 1: Test configuration.


Line differential protection relays are the end-points of the
network connection and are connected to the MPLS routers.
The line differential relays under evaluation included two
types, testing being undertaken for both the trial Ethernet
relays (prototype) and for relays having a conventional X.21
interface (legacy-type). The MPLS routers provide the
Ethernet and X.21 interfaces necessary for connecting the
relays under test and test equipment. Relay test equipment
was used to simulate the current transformer connection and a
conventional lap-top computer was used to extract
management information from the relays. A network tester
was used to generate traffic to simulate the effect of an actual

network and corroborate measurements made by the relays.


An Ethernet Network Emulator was used to introduce
different network profiles (transmission delay, jitter, and
packet loss) to simulate communication burdens and to insert
data errors in the communication paths. Oscilloscopes were
used to simultaneously display the corresponding timing
edges in the relays as shown in Fig. 2.

3.3 Test results and evaluation


a. Ethernet relays
The test cases for the Ethernet relays and the corresponding
measured data are shown in Table 1.
The transmission delay i.e. the latency introduced by an
MPLS router is approximately 20s per router and varies only
slightly. Although the transmission delay increases in
proportion to the number of routers located within a
communication path, this transmission delay is stable. It was
confirmed that transmission delay has minimal effect upon
sampling timing synchronisation and relay operation.
Transmission
delay
[s]
50

Synchronous
accuracy
[s]
-1.0 to +2.0

5073

-0.6 to +2.8

10073

+8 to +20

150 to 187

-24 to +32

Jitter 0.5ms

460 to 700

-120 to +110

Jitter 1.0ms

700

-120 to +120

Transmission
burden 1%
Transmission
burden 20%
Transmission
burden 80%

74 to 77

-1 to +6

79 to 135

-6 to +21

150 to 165

-4 to +12

Test cases

Figure 2: Oscilloscope for measuring the sampling time offset


and jitter between the two relays.
High-performance, high availability multiservice edge routers
are used to deliver differentiated services. A Service
Aggregation Router (SAR) provides IP/MPLS and pseudowire capabilities and is used to groom and aggregate multiple
media, service and transport protocols onto an Ethernet and
IP/MPLS infrastructure. These routers and the Ethernet
Network Emulator were used together to construct the end-toend network experienced by the demonstration relays. To
summarise:
-

Ethernet relays (illustrated in Fig. 3): 3 devices


X.21 relays: 2 devices
MPLS routers: 5 devices (Route via 3 or 4 routers)
Network tester: For performance measurement and
insertion of data errors and communication burden,
packet length: 64/512/1518 byte
Network emulator: Insertion of transmission delay,
jitter and packet loss
Test equipment not shown: For observation and
measurement of latency, jitter and delay symmetry
Relay test equipment: to inject relay input quantities

Normal (via
2 routers)
Transmission
delay 5 ms
Transmission
delay 10 ms
Jitter 0.4ms

Relay
behaviour
(*1)

No
unwanted
operation

*1: Observation of relay behaviour under normal load conditions

Table 1: Ethernet relay test cases and measured data.


However, fluctuation in transmission delay i.e. jitter has a
significant effect upon synchronisation control within the
relays and the accuracy of the sampling timing
synchronisation was reduced.
For the test evaluation for jitter, fluctuations were introduced
in the positive direction using the Network emulator. For
example, in the test case for Jitter 0.4ms shown in Table 1,
jitter is inserted randomly within the range of 0 - 0.4ms in the
communication path from Master to Slave.
The sampling timing synchronisation control function
calculates the synchronous error T and the transmission
delay Td using Equations (1) and (2) below using the similar
mechanism as Network Time Protocol (NTP) and as in
IEEE1588 [4].
Referring to Fig. 4 below:

Figure 3: Fascia of line differential relay showing status


information.

Td = ((T2-T1) + (T4-T3)) / 2.
T = (T2-T1) Td.

(1)
(2)

Master

Slave
T3

T4
T1

SyncReq
SyncResp(T4)

T2
FollowUp(T1)

Figure 4: Sequence for sampling synchronisation


communication.
Where,
- T1: Transmit timing of synchronisation response at
Master side
- T2: Receive timing of synchronisation response at
Slave side
- T3: Transmit timing of synchronisation response at
Slave side
- T4: Receive timing of synchronisation response at
Master side
- SyncReq: Frame for synchronisation request
- SyncResp: Frame for synchronisation response
- FollowUp: Frame for notification of transmit timing
T1
The fluctuation in Td caused by jitter affects the calculated
value of the synchronous error T because sampling timing
synchronisation control is executed using this value
calculated for T. However, prior to finalising the level of
compensation for the synchronous error T generated by
jitter, T is updated by a new delay via the next synchronous
communication. Thus, the actual synchronous errors are
reduced becoming smaller than the jitter introduced.
In the evaluation with regard to transmission burden, the
communication delay is generated according to the length of
the superimposed packets. The network tester was used to
randomly superimpose data packets of 64bytes, 512bytes and
1518bytes. The error in synchronisation accuracy tends to
saturation at a constant value for a maximum packet length of
1518bytes. Regular periods of non-communication are
generated during the confirmation period used to check the
health of the route in the MPLS routers during route
switching. The periods of non-communication are dependent
upon the specifications and settings of the MPLS routers.
When the packets used to simulate transmission burden were
inserted before the packets required for synchronous
communication, the latter are buffered within the MPLS
routers until the transmission burden packets have been
processed and this buffering time is added to the overall
transmission delay. However, a high priority was assigned to
the packets sent by the relays and thus buffering exceeding
one packet was not observed. In addition, the buffering period
was determined by the size of the packets used for the burden
and the transit time and jitter reached a maximum of

approximately 12s (a period corresponding to 1518bytes for


1Gbps communication).
This is different from the jitter test cases in that only a
positive direction in delay can be observed. Therefore the
measured data shows larger synchronous errors for only one
direction. The buffering frequency of the packets used for
synchronous communication increases
under high
transmission burdens, and the number of fluctuations having a
maximum period of 12s increases accordingly.
Currently, the compensation control function has a feature
such that the level of compensation increases for synchronous
errors 10s. Although, as a result, the data recorded in the
test results shows that the fluctuation in the synchronous error
exceeds 12s, the synchronous error was actually stable
within the criteria required for current differential relays.
These test results have highlighted the following two points
as being challenging issues for the trial product.
-

Reduction in sampling timing synchronisation


accuracy caused by transmission delay fluctuation
and policy of 87 relay element block.
Mitigation for non-communication periods in the
event of route switching.

Also, in the evaluation for transmission burden, the sampling


timing synchronisation accuracy is almost stable until
approximately 90% communication burden and the following
insight was obtained.
-

Requirement for a transmission band for protection


relay communications
Calculation of optimal packet length for protection

With regard to route switching in response to communication


failures, SONET/SDH networks can be provisioned to
provide alternate routes for mission-critical traffic such as that
between teleprotection units. When operating correctly,
network reconfiguration can be completed in less than 50ms.
This number is the bar for any new technologies for the
network. In particular, for an IP/MPLS network to be used for
carrying teleprotection traffic, the transmission path must be
restored in under 50ms in the event of a fault. The purpose of
this test was to determine if the network could reroute on a
failure condition and determine the impact on the relays. Here
two failure conditions were simulated. In one case one port
was shut down by a manual command; in the other case the
failure was caused by the manual disconnection of a cable.
The primary route involved 4 routers. With reference to
Fig. 1, the path followed the lower set of IP/MPLS devices.
The secondary route consisted of three routers. The
Network Tester was used to measure the duration of traffic
outage which includes the time taken for the IP/MPLS
equipment to detect the fault and implement the failover
routing plan. The reconfiguration was provisioned to be
revertive, and thus when the fault is cleared, by manual
command or reconnection of the cable, the original route is
reapplied. The route change operation was invoked 5 times to
verify the consistency of behaviour. The IP/MPLS routers
rerouted and reconfigured the network autonomously in the

event of network faults well within the target of 50ms and fast
enough to ensure no detrimental effect in the behaviour of the
current differential relays.
b. X.21 relays
Test cases for the X.21 relays together with the corresponding
measured data are shown in Table 2.
Test cases

Normal (via
2 routers)
Transmission
delay 5 ms
Transmission
delay 10 ms
Transmission
burden 1%
Transmission
burden 20%
Transmission
burden 80%

Transmission
delay
[s]
2777

Synchronous
accuracy
[s]
-12 to -3

7840

-20 to -7

12843

-21 to -6

2903

-28 to -38

2900

-28 to -38

Relay
behaviour
(*1)

No
unwanted
operation

maintaining systems and phenomenon assumption type


WAMPAC applications. Moreover, the introduction of
redundancy, both of the communication route and of the
transmitted data, can result in improvements in the reliability
of the protection scheme. Some examples of expected future
applications are described below.
a. Multi-terminal current differential protection
One of the advantages of current differential protection is that
it can be applied to multi-terminal applications, and it can
perform perfectly in such applications given appropriately
designed communications, as demonstrated by experience
over a long period of time [6]. Where current differential
protection has been applied to circuits of more than three
terminals, it has been most common that a ring
communication architecture has been used, as shown in Fig. 5.

RY1

RY6

RY2

2903

RY3

RY4

RY5

-28 to -38
Figure 5: Multi-terminal ring communication

*1: Observation of relay behaviour under normal load conditions

Table 2: Test cases and measured data for X.21 relay.


In the MPLS routers, serial data for X.21 communication (64
kbps) is converted to MPLS packets at the transmitting
terminal and these MPLS packets are converted to serial data
at the receiving terminal, thus X.21 end-to-end
communication is realized. Communication data is buffered
prior to converting these packets, and the throughput is, in the
main, varied by the setting of the length of the buffer (number
of bytes). This test was carried out using a setting for the
transmission buffer length of 4 bytes and that for the
receiving buffer length of 12 bytes i.e. 1.5ms so that a high
throughput can be obtained. The transmission time delay is
derived from the accumulation of data buffering delays in the
MPLS routers. In the evaluation of transmission burden, we
observed that the transmission delay and sampling timing
synchronisation accuracy were not affected for any of the test
cases.

4 Future benefits
The flexibility and high capacity of Ethernet enables a
number of future benefits to current differential relaying to be
envisaged, such as easier application to circuits of three or
more terminals, the realisation of multi-terminal and widearea back-up protection schemes based on current differential
protection. The technology can support adaptive protective
relaying schemes in which the evaluation of relay setting
margins can be performed based upon the quantitative, online, real-time supervision of operating margins against
varying power system conditions [5]. This approach enables
benefits to be gained from settings adaptations that reflect the
prevailing power system conditions. In addition Wide Area
Situational Awareness schemes using phasor measurement
techniques can be used for predictive dynamic stability

On the other hand, if Ethernet is applied it becomes easier to


send the same data to multiple destinations (or terminals) as a
standard function (multi-cast). In other words, the same
hardware with a single port can be applied to multi-terminal
configurations. The general configuration is shown in Fig. 6.
L2 SW

L2 SW

L2 SW

RY1

RY6

RY2

RY3

RY4

RY5

Figure 6: Multi-terminal current differential protection based


on Ethernet
Needless to say, the relays must be designed so that they can
deal with receiving and processing the data from N-1
terminals within a certain time (where N is the total number
of terminals). In this configuration every relay can calculate
the differential current etc. and operate simultaneously.
b. Wide area current differential back-up system
The idea for a wide area back-up protection system based on
current differential protection has been in existence for a long
time [7, 8], but has never been applied in practice. According
to this theory, the protection system uses the current
differential principle to select the best location for tripping to
occur in order to minimise the area of black out in the event
that the main protection relays or circuit breakers have failed
to clear the fault. This idea has some similarity with the multiterminal current differential protection system, but a
configuration of one central unit and multiple terminal units is
more suitable in this case. Local relays send status data for the
local circuit breakers along with local current data to a central

unit. The central unit is configured so as to accommodate the


configuration of the network to be protected and can decide
the most appropriate circuit breakers to be tripped and send
commands to the local relays. An additional benefit of this
application is that the complex engineering effort normally
required to design back-up protection coverage can be
reduced. Fig. 7 shows the general configuration of wide area
current differential back-up protection.
Central
Unit
L2SW

L2SW

L2SW

L2SW

Local Relays

A S/S

B S/S

Figure 7: Wide area current differential protection based on


Ethernet

be the case in the application as a communication facility for


line differential relays.
However we have confirmed that there is almost no effect
from the number of routers; high accuracy sampling timing
synchronisation is maintained with equivalent performance to
that provided by existing systems using dedicated
communication facilities.
This evaluation has proven that we have reached the point at
which we can now move on to achieve the practical
application of protection relays using a wide-area network.
Furthermore, we have confirmed that exercising the features
already available within communication devices enables us to
enhance the functions and performance of protection relays.
We have also confirmed that existing X.21 legacy relays can
be operated using Ethernet. We believe that this point will
lead to benefits in achieving wider maintainability and
extensibility of the Ethernet protection relays themselves and
the systems in which they are applied.

Two types of system can be envisaged, one type in which the


local relays function only as terminal units for gathering and
sending data to the central unit, and the other type in which
the local relays support independent protection functions and
while only relying on the central unit in case of phenomena
which require a decision based on multi-terminal information.
Various types of network topology can also be considered, the
appropriate system configuration being decided according to
the number of devices required, cost, reliability of
communication, importance of the protected network etc.

In future, we plan to undertake further performance


evaluations using different communication configurations. In
addition, towards the realisation of the practical application of
protection relays compliant with wide-area networks, we plan
to consider and evaluate a protection system which exercises
the performance and features of communication devices such
as communications protocol, security, redundancy control and
priority control.

c. Improvement in reliability by introduction of redundancy


Various types of redundancy can be considered in order to
improve system reliability. In the case of redundancy of the
communication route, data can be transmitted continuously
via two independent routes, and relay operation can be
maintained in the event that a failure occurs in one of the
communication routes. This can be achieved by using Carrierclass Ethernet, which provides a function to assign the
communication route as shown in Fig. 8.

[1] G. Baber, P. Beaumont, F. Kawano, "Current Differential


Protection-over-Ethernet", Cigr Paris 2010
[2] T. Shono et al, "Next Generation Protection System over
Ethernet", IET DPSP 2010
[3] Test Report - Teleprotection over IP/MPLS Networks,
Iometrix 2011
[4] IEEE 1588-2008 Standard for a Precision Clock
Synchronization Protocol for Networked Measurement
and Control Systems
[5] F. Kawano et al, "Intelligent Protection System for Smart
Grid", PAC World Conference 2010 Dublin, Ireland
[6] M. Suzuki, et al, "Present State of Transmission Line
Protection Employing Fiber-Optic telecommunication",
SC34 Colloquium 1987 June, Turk, Finland
[7] Y. Serizawa, and et al., "Wide-Area Current Differential
Backup
protection
Employing
Broadband
Communications and Time Transfer Systems", IEEE
Trans. Power Delivery, vol. 13, no. 4, October 1998, pp.
1046-1052.
[8] J. Tang, P.G. McLaren, "A Wide Area Differential
Backup Protection Scheme For Shipboard Application",
IEEE Transactions on Power Delivery, Vol. 21, No.3,
pp.1183-1190. (July 2006)

L2 SW
Ry-A

Route 1

Ry-B

Route 2

Figure 8: Redundant communication based upon Carrier-class


Ethernet

5 Conclusions
Prior to the tests, we had envisaged that the accuracy of
sampling timing synchronisation would deteriorate in
proportion to the increase in the number of MPLS routers
within the communication path. We had also envisaged that
constraints in network configuration and system scale would

References

You might also like