500 1

Investigations of different Function Allocations

in SA Systems enabled by IEC 61850
Klaus-Peter Brand, Senior Member, IEEE, and Wolfgang Wimmer

access to the data is open by standardized communication

Abstract— At least since 2006 the standard IEC 61850 for services. Both together and the allocation of LNs to the
communication in substations is in commercial use providing switchyard elements (single line) is described in the
interoperability in Substation Automation systems between Substation Configuration description Language (SCL) for
devices from different suppliers. Its approach is based on the interoperable exchange between engineering tools. The data
split between applications and communication, functions and
and communication model is mapped to a main stream
implementation and, therefore, supports also the free
allocation of functions to IEDs. This allows the question for an
communication stack consisting of MMS, TCP/IP and
optimal allocation regarding the functional requirements of Ethernet [2].
SA, its impact on the communication system, performance of The allocation of functions in terms of LDs and LNs to
functions, engineering effort and reliability, and on the cost
physical devices (IEDs) may have a strong impact on the
level. Common SA systems are some mix between a total
centralized and decentralized allocation providing the well-
communication architecture, on reliability, investment cost,
known functions on station, bay and process level. The two and engineering effort and maintenance of the SA system.
extreme solutions are positioned and it is evaluated if the Especially, the reliability has also an impact on the overall
common mix is an optimal one. The result is a proposal for the network management system, and therefore on the reliability
best allocation of existing functions to IEDs. Depending on of power supply in general.
given functional requirements and not too strict boundary
conditions, improvements are possible exploiting the full Up to now, functions in SA systems are allocated in a very
capabilities of IEC 61850, if accepted. traditional, fixed way to IEDs. These fixed devices are
normally applied on bay level with an interface to the
Index Terms – Substations, substation automation, switchgear on process level and others to HMIs at station
communication, communication architectures, protection, and bay level, and to a gateway to the network control
control, IEC standard, IEC 61850, function allocation, center (NCC) on station level (see Figure 1). In case of
reliability, performance, optimization, centralized and serial links they are called process bus and station bus.
decentralized systems There are some minor deviations, e.g. combined IEDs for
protection and control on MV or distribution level, and the
I. INTRODUCTION disturbance recorder function built into most protection
Substation automation (SA) is commonly used in IEDs today. There is the implicit assumption that this
controlling, protecting and monitoring of substations [1]. Up approach is the optimal one. The standard IEC 61850 allows
to now, the communication for SA has used private serial many other allocations and related communication
communication systems complemented by conventional architectures. An evaluation of the reliability of different
parallel copper wiring, especially between the process communication architectures enabled by IEC 61850 was
(switchgear) and the bay units. With the finalization of made already in [3].
IEC 61850 [2] in June 2005, there is a comprehensive Sequences
global standard for all communication needs in the Archiving
Interlocking Station
Station
substation being introduced now. The standard provides Level
HMI BBP Computer
Gateway

interoperability of devices from different vendors, which

Station
has been proven not only by demonstration and pilot Bus

systems, but also by a lot of commercial substation

automation systems set in operation mainly in 2006 and
beyond. Another important feature supported by IEC 61850 Bay Protection Control &
Protection Control &
Level Monitoring Monitoring
is the free allocation of functions to physical devices (IEDs)
which is addressed in this paper. Process Process
Bus Bus

The approach of IEC 61850 separates application and

communication. The application is represented by the object
oriented data model with standardized basic function atoms Sensor Actuator Sensor Actuator Sensor Actuator Sensor Actuator

called Logical Nodes (LN) classes which contain the Process

Level Bay Bay
function related standardized data. LNs can be grouped for
Switchgear
managerial and organizational purposes in to Logical
devices (LD) which are hosted by physical devices. The

The authors are with ABB Switzerland Ltd., Baden: Figure 1 – Common structure and function allocation in SA systems
klaus-peter.brand@ch.abb.com, wolfgang.wimmer@ch.abb.com
500 2

II. THE FREE ALLOCATION ACCORDING TO IEC 61850 their functionality. The station level is consisting of an HMI
IEC 61850 provides not only interoperability, but supports with work place functionality only and of a gateway to the
also the free allocation of LNs to physical devices to be NCC. The IEDs can be physically clustered as optimal
open for different design and operation philosophies of the needed from reliability or maintenance point of view. Also
utilities world-wide. Important for the future is that this this solution is supported by the powerful communication
approach opens the door for an assessment of given SA standardized by IEC 61850 providing the independence of
solutions and to ask for the optimized solution proofing the the data model and communication services from
existing traditional ones or claiming for new ones based on implementation. Any single IED failure refers only to one
the well known functions needed by the utilities in SA atom of functionality. The influence of a function atom
systems. This new degree of freedom enabled by IEC 61850 failure on the functions must be minimized by an
may result also in developments extending the functionality appropriate design (see Figure 3).
of the substations and SA system for the benefit of the
overall power system. It should be noted that any packaging
Station
HMI
of functions has an impact on the SW implementation and, Level
Gateway

therefore, is within the responsibility of the suppliers Station

responding to market needs. Bus

Interlocking
Interlocking
Archiving Sequences Archiving
Sequences
III. BASIC EXAMPLES OF FUNCTION ALLOCATIONS Bay Protection Control Monitoring Protection Control Monitoring
Level
One extreme example is the allocation of all functions BBP BBP

needed in the substation to one central computer at station Process

Bus
Process
Bus
level collecting with a vast number of sensors at process
level all data needed, and is issuing any actions like Protection Protection

commands and trips via remote outputs (actuators) at Sensor Actuator Sensor Actuator Sensor Actuator Sensor Actuator

process level again. There are no bay level and no dedicated Process
Level Bay Bay
IEDs for protection, control, monitoring and other functions
Switchgear
anymore. Since the necessary computation power was in the
past available in big computers only, such solutions had
been proposed at the very early beginning of substation Figure 3 – Toward a totally decentralized SA system (schematic)
automation. They may now see a revival because of the The most SA systems today have populated all three levels
powerful communication services standardized by with functions in multifunctional IEDs. At station level, the
IEC 61850. There is no discussion that such a centralized computer is not only providing an HMI, but also functions
computer has to be duplicated to avoid a very dangerous for the complete substation like event and alarm lists, an
single point of failure (see Figure 2). The loss of sensors archive for historical data and e.g. functions like a
and/or actuators is a problem left for the adaptability of sequencer supporting the local or remote operator. At bay
functions facilitated by the comprehensive data image in level, dedicated IEDs for functions like protection, control,
these centralized computers. The bay level control has to be and monitoring exist. With the use of a serial link (“process
connected also to these central computers, making it more bus”) according to IEC 61850 between the bay and process
vulnerable against cable faults in the system. level, also at process level IEDs with sensor or actuator
functions may be found (see Figure 1). Multiple measures
are taken to overcome single point failures and to increase
Sequences
Archiving
reliability. Therefore, the common solution with process bus
Station
HMI
Interlocking Station
Gateway
looks like an intermediate one mixing benefits and
Level BBP Computer
Protection drawbacks of both extreme ones. The paper tries to answer
Control
Station
Monitoring
the question for the optimal solution regarding the
Bus
functionality requested in the substation, and all related
Process
Bus criteria.
Bay
Level IV. EVALUATION CRITERIA

A. Load on the communication system

The load on the communication system is different for
Sensor Actuator Sensor Actuator Sensor Actuator Sensor Actuator different architectures as result from different function
Process
allocations. Especially, the load of the communication
Level Bay Bay interface per IED is strongly dependent on this allocation.
Switchgear
Since the functions including their data exchange have to be
performed with the same performance independently where
Figure 2 – Totally centralized SA system with sensors and actuators they are implemented, the needed performance of the
IEC 61850 implementation may restrict the free allocation
in some cases.
The other extreme example is the total distribution of all
functions, maybe down to one LN per IED. The related
IEDs will be both on bay and process level depending on
500 3

B. Reliability the complete substation. Different allocations of the

The reliability of communication depends on the resulting included functions to IEDs will result in more or less
communication architecture with all links and optimized solutions.
communication components like switches. The reliability of
different communication architectures based on IEC 61850 B. Local protection functions
was evaluated already in [3]. The allocation of ‘function Bay local protection functions are functions which need
atoms’ to IEDs influences the physical architecture needed only voltages and currents from one bay and act in any case
for the function implemented and, therefore, the reliability on the breaker in this bay only. There is a weak informative
of the function. Especially, more IEDs results in more link e.g. to the HMI. If the voltage and current samples are
switch ports and therefore in more or bigger switches. coming from different places over the process bus, the
behavior is getting non-local and more depending on the
C. Engineering effort
allocation.
The number of IEDs for the allocated LNs, the complexity
of the functions per IED, and the complexity of C. Distributed protection functions
communication increases the engineering effort and the
Some protection functions need information from many
power needed by the system engineering tool. The
bays or act on different breakers by definition like busbar
IEC 61850 approach of decoupling functions from the
(BB) protection or breaker failure protection. Such
physical IEDs and the communication architecture can again
functions behave sensitively to their allocation to IEDs.
decrease the engineering effort. The semantic included in
the definition of the Logical Node classes and their Data
Objects supports even automated function engineering or at VI. EVALUATION
least consistency checks. If such an approach is fully used, In the following only configurations with process interfaces
the engineering effort for all kinds of solutions with same to the switchyard are considered. This interface is always
functionality will be decreased and in the same order. located in the switch gear and a process bus used.
D. Costs A. The extremely centralized case: Benefits and
Detailed costs of a SA system are depending on the drawbacks
technology and the market approach of the provider and the The totally centralized solution as given in Figure 2 consists
competition on the market. In general, based on the of a redundant functional node (protection, control,
application different classes of IEDs may be classified by monitoring, etc.) at station level interacting with all sensors
cost tags based on the computation requirements. It may be and actuators of the SA system at process level constituting
assumed that sensors and actuators work only with GOOSE the remote process interface.
and SV services. Devices with exactly one functional LN
are supporting all services. Similar today, IEDs containing 1) Load and Performance
groups of about 20 LNs have also to support all services but The assumption for the communication load calculation is
need also higher computation power due to more functions that each voltage (LN TVTR1) and current transformer (LN
performed and data communicated. Finally, the highest TCTR1) is producing sampled value (SV) messages with
processing power for both functions and communication is sample rate of 4000 Hz and length of 1 kBit per message,
needed by the central IEDs containing the LNs of all and that each switchgear (LNs XCBR1 or XSWI1) and the
functions needed in the complete substation. A very rough centralized protection and control is producing one GOOSE
cost estimation results in the following data in Tab. 1. message/s (msgs/s) with the length of 10 kBit each.
Commands are contributing same as GOOSE messages.
IED Class Description Relative Cost TVTR1 GOOSESAV BufferedUnbuffered FuFail
Volt
FuFail
Volt
PB
SAI Sensors and actuators 0.2 TCTR1 Current Current

process bus (PB) interface PB

XCBR1 Pos Pos
LNA Single function LN 1 PB
BU Multi function LNs 2 QB1XSWI1 Pos Pos

(about 20) PB
QB2XSWI1 Pos Pos
CU Central LN processing 20 PB
(all – e.g. 400 - LNs) QB9XSWI1
PB
Tab. 1 – Relative IED costs QC1XSWI1 Pos
PB

V. FUNCTION IN SA SYSTEMS (EXAMPLES) QC2XSWI1 Pos

PB
QC8XSWI1 Pos
A. Distributed control functions PB

Since already simple commands depend on the sequence of Operate CentralProt

PC
PB
steps between the HMI at station level or in the remote Operate Operate Operate Operate Operate Operate Operate CentralCntl
NCC, the bay controller and the process interface, there are PC
PB

no real local control functions. In most cases, commands are

checked against interlocking and synchrocheck conditions
arising from relevant subparts of the substation or from the Figure 4 – Central solution with 1 bay connected
complete one. Automatic sequencers may be used acting at
500 4

In Figure 4 the communication in a process bus (PB) type protection (Prot1) to the station level (LN IHMI1, LN
network between the central combined protection and ITCI1). If applicable the fuse failure from the voltage
control unit and the sensors and actuators for one bay with 2 transformer (LN BB1_TVTR1) is also reported. The trip of
instrument transformers and 7 switching devices is shown. protection is sent by GOOSE message to the control in the
The resulting load per such a sub-network is 8090 kBit/s or same bay e.g. for autoreclosure start. The positions of the
8009 msgs/s. The throughput limit given by the actual switchgear are exchanged between bays represented here by
version of IEC 61850 is 100 Mbit/s per line. Switches Ctl1 and Ctl2 with GOOSE messages for interlocking. Same
having also ports with 1 GBit/s are available and mostly happens for the positions from Ctl1 to Prot1. If applicable
used for backbone links. To avoid collisions on the Ethernet the voltage of the busbar VT (BB1_TVTR1) is send as SV
and support priorities as requested by IEC 61850, managed message to the bay controller (e.g. Ctl1) for the
switches have to be used which have a throughput in the synchrocheck function.
order of 200’000 msgs/s. Therefore, the calculated load for The resulting load per bay in such a sub-network including
one typical bay means that such a solution with centralized the voltage samples is 4038 kBit/s or 4011 Msgs/s. The
functions can handle around 10 bays per sub-network or main load comes from the busbar VT since the calculation
network segment related to a 100 MBit/s link to the central without this VT results in 36 kBit/s or 9 msgs/s.
IEDs, and around 20 bays related to the switch message This means, that if the bay level samples can be hidden from
throughput. Due to the needed high amount of input the interbay bus (IBB), at least 1000 bays can be served by
processing capacity, especially for samples (SV messages), the sub-network without distributed synchrocheck or,
such a group of 10-20 bays may need a separate sample pre- generally, without broadcasting samples of current and/or
processing unit within the centralized function unit. voltage. In this case, the real bottle neck is the throughput
2) Loss of sensors and actuators into the station level IEDs. If this would be 1000 msgs/s,
Regarding the effort for sensors and actuators and the around 200 bays are possible. The synchrocheck adds
communication load, in most cases (except complete around 4 MBit/s, i.e. 1 Busbar VT corresponds to the load
separation of main1 from main2 in HV protection) these of 100 bays without sampled analogue values (SAV). If we
components will not be duplicated. Therefore, the most stay with 100 bays resulting in maximum to loads of about
important drawback of such a solution is the impact of the 4 MB/s, roughly 20 VTs could be handled in addition. This
loss of a sensor or an actuator or of the related then relates either to a maximum of 20 bays or 20 bus bar
communication link. The MTTF of sensors and actuator segments depending where the busbar VT values are created
may be assumed to be at least 200 years, SA proof switches and how communicated. This means that in the
as far as needed have currently around 50 years. The 'conventional' architecture with BB VT values onto the IBB
centralized common data base of the complete substation around 20 HV bays could be handled. Also 20 bus bar
offer the basis to compensate some losses, but innovative segments with totally up to 100 bays seem to be a
new algorithms are needed to handle this problem. reasonable size, without technical problems. If for medium
voltage (MV) the sampling rate may be lower, this figures
B. The extremely decentralized case: Benefits and could be even higher. The critical factor then becomes the
drawbacks input processing capacity of the IEDs.
The overall functionality of the totally decentralized There is also a direct interaction between the bays including
solution as indicated in Figure 3 consisting of many the exchange of current samples for busbar protection,
functions or function groups is implemented by which is not considered above; for MV systems it is seldom
communication between many decentralized IEDs at bay used, for HV systems it is normally a separate network.
level, each hosting one LN. .
IHMI1 GOOSE SAV Buffered Unbuffered
PC
1) Load and Performance IBB
The assumptions about the messages generated are the same ITCI1
as above but the results are different by the different PC
IBB
topology. What is shown in Figure 4 as CentralProt and Tripevents Tripevents Prot1 Trip
CentralCntl now becomes a single function processor, e.g. Prot
IBB
for instantaneous overcurrent protection (PIOC), for
BinState BinState Pos Ctl1 Pos
synchrocheck (RSYN) or for switch control (CSWI), and is Measurents Measurents BC
multiplied up to one IED per LN needed for the function IBB

considered. These LNs add a few GOOSE messages for Measurents Measurents Pos Ctl2
BinState BinState BC
communication in between them, and some reports for IBB
supervision and operation purpose to and commands from FuFail FuFail Volt Volt BB1TVTR1
PI
station level. If we assume typical the implementation of IBB
20 LN instances of different LN classes for bay
functionality, this adds around 40 msgs/s resp. 40 kBit/s. Figure 5 – One conventional bay exchanging Interbay data
Report messages from a bay unit both to the IHM and NCC and receiving samples from busbar VT
gateway may assumed to contribute with a message size
same as the GOOSE messages, and sent once per second.
Figure 5 illustrates the load between the bays and to station
level. Commands are sent from the HMI (LN IHMI1) and
NCC gateway (LN ITCI1) at station level to the bay control
(Ctl1). Reports are sent from the bay control (Ctl1) and bay
500 5

2) Loss of single functions or function groups IED) needs careful configuration of the communication
This decentralized solution is very sensitive to the loss of a system to limit the data flow. The demand for processing
single LN involved in multiple functions, caused either by power is more on the communication side than on the
the loss of the allocated IED or of relevant communication functionality side. Therefore, this solution is not very cost
links. The station level consists of the HMI only. Therefore, effective. If we apply the estimation of Tab. 1 to system
a certain redundancy of functions as already common at HV sizes of 5, 10 and 20 bays each with 20 Logical Nodes, the
level as main1 and main2 protection or as back-up system hardware costs relate as follows (sensors and
protection may be requested in general. It is mainly needed actuators neglected because they are the same in all cases).
for single LNs acting as interface to the process or for
functions being commonly used by more than one other
function like autoreclosure (represented by RREC)and the Cost/Case Distribute Conventiona Centralized
trip conditioning matrix (represented by PTRP). For d l
communications redundant or duplicated systems may be 5 bays 100 10 40
necessary and useful [3]. 10 bays 200 20 40
20 bays 400 40 40
C. Today’s allocation: an optimal solution?
Tab. 2 – Cost comparison of systems
• Case 1: Station and process bus physically separated
The most common SA architecture shown in Figure 1 is The today’s bay oriented solution shows a reasonable
based on the allocation of functions to some more or less function allocation (some functions per IED), is best
dedicated IEDs. All IEDs have an interface to the process adaptable to system size, and costs are never higher than the
data of this bay, and a second interface to the station bus centralized solution. It allows the maximum system size
communication to the station level and between the bays. with reasonable reliability and reasonable communication
• Case 2:Common bus for station and process system engineering effort (see Tab. 3). Similar
communication
improvements in reliability as in the central case against loss
The bay level functionality is allocated to bay level IEDs
of sensors with relatively low costs may be possible by
which exchange data by one interface and one switch both
for process bus data and interbay and station data as having a central IED per busbar section and function group
described in [3] as solution in Figure 7. (Control, Protection) instead of per complete system. The
1) Load and Performance data flow between busbar sections and system level is
Case 1 separates station bus traffic from process bus traffic minimized and the functional redundancy can be used to
by architecture. The critical point to be solved is how some some degree. To exploit this solution the development of
process data needed in other bays can be brought to the some new distributed algorithms is needed. The relative
station bus side (e.g. voltages for Synchrocheck). positioning regarding cost may be influenced by future
Case 2 has not this problem, but the separation must be advances in technology and could be even lower than for
handled by careful configuration of the switches e.g. by the conventional solution.
VLANs. This separation assures that the very high amount
of about 200 bays can be handled. Case Distributed Conventiona Centralized
2) Loss of components l
If we loose one protection IED (e.g. main 1), there is at least Maximal 100 - 200 200 10-20
on transmission level a second protection (main 2) available. number
It should be noted that main1 and main 2 have some other of bays
purpose beyond redundancy (diversity in fault detection
Reliability Only single Loss of one Compensate
characteristics). The MTTF of control IED is mostly
function bay tolerated loss of one
sufficiently high, and the loss of one bay regarding control
is in most cases acceptable, especially if there is an concerned sensor;
emergency device directly at the switchgear. bay HMI
The loss of sensors and actuators leads to loss of the part of vulnerable
the functions. Critical are the sensors both for current Switch High Low to Low
(TCTR) and voltage (TVTR), and the circuit breaker configura- (VLAN medium
interface XCBR. The loss of the circuit breaker can be tion effort config..) (VLAN
handled for protection purpose by the breaker failure config.)
protection. Tab. 3 – Comparison of system parameters

VII. CONCLUSIONS A general architecture like the conventional one, combined

The fully centralized solution is limited both by the with communication architecture as suggested in [3] Figure
communication and processing capability. The risk to be 8, seems favorable at the current state of technology. With
dependent from the central configuration tools of one decreasing processing costs some degree of centralization
supplier is high, and no retrofit or expansion with other for some functions, e.g. distance protection per bay, bus bar
solutions is possible. A bay level HMI is more vulnerable protection, control per station, might become more cost
against communication system faults, because it is also effective, but they mostly need some development and,
connected to the central IED. more hindering, an acceptance by the utilities as users. In
The fully decentralized solution (one function atom per any case the roadmap to future systems is given. To allow
500 6

approaching the optimal function allocation by any future

retrofit or add-ons as well as having minimal engineering Klaus-Peter Brand (SM’89) was born in Neustadt/Aisch, Germany, in
1948. He studied Physics in Würzburg, Kiel, and Bonn (Germany). He got
effort, it is recommended to always model each SA system his Master (Dipl.Phys.) and his PhD (Dr.rer.nat.) from the University of
at the LN / LD level as if it would be completely distributed. Bonn. 1976, he joint the plasma physics group (SF6) of BBC/ABB
Any dedicated project may have all solutions at disposal, Research Center in Baden, Switzerland. From 1982, he was in different
positions strongly involved developing substation automation systems and
but may be limited by some boundary conditions. Therefore,
building up this business in ABB, Switzerland. He is working at the ABB
the best system design remains still a challenge for the University Switzerland as instructor and consultant. He is engaged in
skilled substation automation system engineer. CIGRE SC B5. From 1995, he is being member of the AHWG and WG10
of IEC TC 57 worked from the beginning defining the standard IEC61850.
He is acting now as expert, editor and co-editor of parts of this standard for
REFERENCES maintaining and creating the 2nd edition. He is also chair of the Swiss
[1] K.P.Brand, V.Lohmann, W.Wimmer “Substation chapter of IEEE PES.
Automation Handbook”, UAC 2003, ISBN 3-85759-
Wolfgang Wimmer, was born in Bad Schwartau, Germany, in 1947. He
951-5, 2003 (www.uac.ch) studied Mathematics and Computer Science at the University of Hamburg
[2] IEC 61850 “Communication networks and systems in (Germany), and also got there his Master (Dipl..Inf) and his PhD
substations”, 2002-2005 (www.iec.ch) (Dr.rer.nat.). In 1979 he joined BBC/ABB in Baden, Switzerland. From
[3] L.Andersson, K.P. Brand, Ch.Brunner, W.Wimmer 1983, he was in different positions strongly involved developing substation
automation systems and building up this business in ABB, Switzerland. He
“Reliability investigations for SA communication
is working presently at ABB Utility Automation, Switzerland, as principle
architectures based on IEC 61850”, IEEE PowerTech systems engineer in the development of substation automation and
2005, St.Petersburg, Paper IEEE_PPT05-604 monitoring systems. From 1996 he is a member of IEC TC57 WG10
working on the standard IEC61850. He is acting now as editor and co-
editor of some parts of this standard.