Documente Academic
Documente Profesional
Documente Cultură
net/publication/224400747
CITATIONS READS
8 4,015
3 authors, including:
Some of the authors of this publication are also working on these related projects:
EV in Smart home energy management: reliability and economic acomplishment View project
All content following this page was uploaded by Akhtar Kalam on 29 May 2014.
Abstract-- In the past decade, new communication schemes have (3) The TimeClient, then, computes the difference between
been designed and retrofitted into substations by utilities to its own time setting and the time setting it receives from
integrate data from relays and Intelligent Electronic Devices the TimeServer and uses this difference to adjust its
(IEDs). The idea of standardising the language of communication internal clock bringing it into phase with that of the
between IEDs has evolved as the key for the advancement of
connectivity and interoperability within a substation automation
TimeServer.
system. As a result of the standardization process, one very
critical communication standard has evolved, the International
Electrotechnical Commission (IEC) 61850. IEC 61850 describes
how devices are to communicate in a substation as well as the Time request from client to server
related system requirements. Time Synchronisation (TS) is the
Client1 Time Server
synchronisation of the date and time of all devices in a network. Time response from server to client Switch
It is crucial in time-sensitive substation applications and its
importance has further been clearly acknowledged in the IEC
61850 standard as a requirement. As part of an undertaken
project, an object-oriented implementation of the IEC 61850
standard as a concrete application layer protocol was carried out
and a message-oriented middleware architecture was proposed Client2
as a means of integrating the various functionalities that support Figure 1. The basic TS process
the unique behaviour and communication needs of the IEC 61850
standard. This paper focuses on the implementation of a The accuracy of the TS depends heavily on the TS protocol
commercial off-the shelf TS protocol and its incorporation into used as well as the performance of the underlying hardware.
developed communication model.
The delays encountered by a TS message from the moment it
Index Terms— ACSI, Communication Systems, IEC 61850, IED, is prepared in the TimeServer until it is executed in the
Protocol, Time Synchronisation, Standardization, Substation. TimeClient are the main sources of TS error. The two most
concerning delays are the communication stack and network
I. INTRODUCTION transmission delays. Network Time Protocol (NTP) [1] is
regarded as the most accurate and flexible means of
Time Synchronisation (TS) is a crucial element of network synchronising clocks over the Internet and across Local Area
design and implementation. A time-synchronised network is Networks (LANs) with an accuracy of a few milliseconds
vital for the operation of network applications with optimal (ms). TS is also critical in sensor networks where applications
performance. The ultimate goal of TS is to bring the local such as power system protection and control require collective
clocks of servers and other instrumentation in a network into processing of time-sensitive data. In such applications, sensing
phase so that their time differences will be zero. A typical TS and actuation need to be coordinated across multiple nodes
process, shown in Fig. 1, may be divided into the following [2].
steps:
The new IEC 61850 [3] is a Substation Automation (SA)
standard developed by the IEC Technical Committee (TC) 57.
1
(1) One of the nodes within the network is chosen as the time
server (TimeServer) signifying that all other nodes within The model-driven approach of the IEC 61850 standard
the same network need to synchronise their clocks with describes the communication between devices in a substation
the TimeServer’s local clock, and the related system requirements. It supports all substation
(2) A time client (TimeClient) initiates a time request to the functions and their engineering by using Object Oriented (OO)
TimeServer receiving the TimeServer’s local clock time data models that describe the processes to be implemented and
within a reply packet, controlled, e.g. the functionality of a circuit breaker or a
feeder equipment etc. As a result, it standardises the language
of communication between devices in a substation automation
1 system allowing for the free exchange of information.
C.R.Ozansoy is currently working as a lecturer in the School of
Engineering and Science, Victoria University, Melbourne - Australia (e-
mail: cagil.ozansoy@vu.edu.au). Although the IEC 61850 set of documents consists of 10 parts,
Aladin Zayegh is a senior lecturer in the School of Engineering and the most important contents are found in Parts 7-x [4 - 7]. Part
Science, Victoria University, Melbourne - Australia. 7-2 [5] specifies the basic layout for the definition of the
Akhtar Kalam is a professor in the School of Engineering and Science,
Victoria University, Melbourne - Australia. Abstract Communication Service Interface (ACSI) object and
service models. Part 5 [8] specifies the need for TS amongst
2008 Australasian Universities Power Engineering Conference (AUPEC'08) Paper P-012 page 1
the devices of a SA system. The components of the TS model, II. SIMPLE NETWORK TIME PROTOCOL
as specified by the IEC 61850 standard, are shown in Fig. 2.
SNTP is a simplified version of NTP containing only a subset
Time synchronisation
IEC 61850 Time of the NTP functionality [9]. SNTP lacks some of the internal
IEC 61850 Client/ Server
Server
algorithms of NTP such as the advanced filtering techniques
Time request (Master UTC Time) used to control variable latency. However, it is still considered
(UTC Synchronised to be adequate in meeting the TS demands of many systems
Time) * externally
synchronised
within acceptable accuracies.
Time response
As a part of a PhD project, the authors focused on developing • Reference Timestamp: Time at which the local clock is
an object-oriented implementation of the IEC 61850 standard last set or corrected in the TimeClient,
as an application layer protocol running on a middleware • Originate Timestamp (T1): Time at which the time
architecture specifically designed and implemented to support request leaves the TimeClient headed to the TimeServer,
all the communication details required by the standard. In this • Receive Timestamp (T2): Time when the time request
paper, the focus is on describing how a fine-grained TS arrives at the TimeServer,
function can be incorporated into the designed communication • Transit Timestamp (T3): Time at which the time reply
architecture such that the clocks of all devices within a leaves the TimeServer headed to the TimeClient, and
substation network can be synchronised. • Destination Timestamp (T4): Time when the time reply
arrives at the TimeClient.
The objective in this paper is to demonstrate if the SNTP TS
protocol, with additionally included features, would be The sequence of events that take place in a unicast SNTP
sufficient to meet the accuracy requirements in substations by application is as follows:
time stamping at the lowest possible stack layer instead of the
application layer. SNTP has been chosen for implementation a) The T3 field of the time request is set to the time of the
due to its relative simplicity and as it is specified in the day by the TimeClient according to its own local clock,
standard [8]. This paper only outlines the synchronisation of b) When the time request arrives at the TimeServer, it copies
substation nodes relative to a time-server excluding the the T3 field into the T1 field and further sets T2 and T3
process of external synchronisation of the time-server’s local fields according to its own local clock before forwarding
clock relative to a real-time clock. A possible future inclusion the time reply back to the TimeClient,
would be the modelling and implementation of a GPS device c) When the time reply arrives at the TimeClient, it
to take into account the process of external synchronisation. determines the time of its arrival and sets this time into
the T4 field,
2008 Australasian Universities Power Engineering Conference (AUPEC'08) Paper P-012 page 2
d) The client application calculates the RTD, LCO and • Transmit protocol stack delay: is the delay a packet
dispersion based on (1), (2), and (3) respectively [1]: encounters when travelling from the application layer,
where it is assembled, all the way down to MAC layer.
RTD = δ = T4-T1-T2+T3 (1) Also referred to as “Send Time”, this delay is highly non-
LCO = θ = (T2-T1+T3-T4)/ 2 (2) deterministic,
Dispersion = ε = ρ + ϕ(T4 − T1) (3) • Receive protocol stack delay: is the time it takes an
incoming packet to travel from MAC layer up till the
Where T1, T2, T3, and T4 are the timestamp values, ρ is the application layer. Also referred to as “Receive Time”, this
measurement error, and φ is the skew rate given by (4): delay is also highly variable, and
• Switch latency: is the time it takes a networking device
SNTP.MAXSK EW such as a router to process a data frame. Switch latency is
ϕ= (4) highly dependent on the architecture of the networking
SNTP .MAXAGE device and also the network load. The amount of time a
packet is delayed through a switch may vary differently
Where MAXSKEW and MAXAGE are constants denoting the based on the switch load.
maximum skew error and maximum clock age respectively.
Skew is the frequency difference between the local and Although SNTP does not actually incorporate any filtering
reference clocks. Clock age is the duration a reference clock is mechanisms, the need for such a mechanism to deal with the
considered valid [1]. The client application uses the value of variable switch latency was clearly recognised in this project.
the calculated LCO to advance its local clock with t + LCO Hence, an important feature of NTP, the clock filter procedure
units during the next t time units bringing it into agreement taken from [1] was also incorporated into the SNTP model
with that of the TimeServer. The client application can implemented in this study. The NTP clock filter procedure
calculate the required frequency (f´) using the formula: f´ = f / uses the clock offset, roundtrip delay and dispersion variables
(1-LCO/T) where f is its current frequency and T is the as its input arguments. It is executed each time a new NTP
synchronisation period [11]. message arrives (in this case SNTP) where a new set of data
samples (θ, δ, ε) is calculated and shifted into the filter at the
In a publish/subscribe SNTP application, the TimeServer uses left end. A shift register consisting of many shift stages is used
a multicast group address for periodically forwarding time for the storing and shifting process. The basic idea of this
updates. The key disadvantage concerned with the procedure is to calculate the filtered clock offset, roundtrip
publish/subscribe mode of operation is the fact that delay and dispersion values updating the dispersion of the
TimeClients can not calculate the RTD based on equation (1) samples previously received and saving the current time. It is
since the received multicast messages only have their T1 and based on the computation of a quantity called the
T4 fields set. However, this disadvantage can easily be synchronisation distance (λ) from the roundtrip delay and
eliminated if each TimeClient sends a single dummy unicast dispersion using:
packet to the TimeServer at the start-up with the intention of
calculating the RTD. When the dummy unicast packet is
|δ |
received back from the TimeServer, it will have all the λ =ε + (6)
timestamps making it possible to calculate the RTD. Once the 2
RTD is calculated, the TimeClient only listens and waits for
the multicast time updates. On their arrival, the client All sets of samples contained within the filter are sorted by
applications simply calculate the LCO and make the necessary increasing synchronisation distance and the set of values with
adjustments based on (5). the minimum synchronisation distance is chosen as the end-
products that correspond to the filtered clock offset, roundtrip
LO_multicast = (T1-T4) + (RTD/2) (5) delay, and dispersion.
In this study, a collection of freeware Microsoft Foundation III. IMPLEMENTATION OF SNTP CLIENT AND SERVER
classes taken from the CSNTPClient project [12] were used APPLICATIONS
for implementing the core SNTP structure
SNTP client applications can be integrated into IEC 61860
B. SNTP Filtering applications running at the ACSI application layer modules of
the communication processor as shown in Fig. 3. They share
This section outlines the use of adaptive filtering techniques the same connectivity functionalities of the underlying
for coping with the sources of errors in TS schemes. The middleware for interacting across the network. An ACSI
major cause of error in a TS scheme, where Ethernet is being server represents the external visible behaviour of a real
used as the transfer media, is the non-deterministic structure of device and acts as a SNTP TimeClient. In any network, there
the Ethernet. This non-determinism results in variable could be as many as SNTP clients. However, hypothetically a
message delivery delays. The major sources of TS error, all single SNTP server is allowed. In the study outlined in this
stemming from the non-deterministic property, are reviewed paper, the SNTP server was designed and implemented in a
below [2, 13]: separate communication processor as a single-application
running node as shown in Fig. 4.
2008 Australasian Universities Power Engineering Conference (AUPEC'08) Paper P-012 page 3
A. Application Layer Process Modelling of a SNTP Client
SNTP Client IEC 61850
application Server application The State Transition Diagram (STD) of the entire application
layer module of an ACSI server is shown in Fig. 5. The intact
state diagram consists of six states, only two of which are
related to the SNTP application. They are the “Time_Syn” and
“Decide” states. The remaining states are all related to the IEC
61850 related applications occurring within the server.
Is
packet
No from a Yes
Multicat
SNTP
server?
Obtain all four Obtain the
timestamp values Transmit and
Receive
Timestamp values
Compute the LCO
,RTD and skew
Compute the LCO and skew
End
2008 Australasian Universities Power Engineering Conference (AUPEC'08) Paper P-012 page 4
The SNTP client application is capable of supporting both the
unicast and multicast modes of operation. In both cases, the
necessary timestamps are acquired, and LCO and skew are
calculated for use in the process of correcting the local clock
relative to the chosen TimeServer.
2008 Australasian Universities Power Engineering Conference (AUPEC'08) Paper P-012 page 5
the TimeClients and TimeServer further imposed additional
jitter worsening the variable message delivery delay times.
These sources of errors deteriorated the TS accuracy as shown
in Fig. 10 where it is clearly demonstrated that only the IEC
class T2 timing accuracy of ± 0.1ms can be achieved in such a
network without the use of any filtering mechanisms.
Therefore, the need for the use of filtering mechanisms in
eliminating the effect of variable switch latency was once
more comprehended.
B. 5% load case
2008 Australasian Universities Power Engineering Conference (AUPEC'08) Paper P-012 page 6
SNTP protocol sufficiently assists in the attainment of the TS
accuracy requirements set by the IEC 61850 standard in a low
data exchange network by a means of time stamping at the
MAC layer and with the use of an adaptive filtering technique.
2008 Australasian Universities Power Engineering Conference (AUPEC'08) Paper P-012 page 7