1

A Quantitative Evaluation of IEC61850

Process Bus Architectures
Jean-Charles Tournier, Thomas Werner

Abstract—The IEC 61850 standard plays an important role in

Network Control Center
the smart grid landscape by providing an unified way of accessing
and exchanging data related to substations equipment. While Engineering
Station PC Workplace
large parts of the standard are widely adopted and integrated Gateway

by the different vendors, the 9-2 part, or process bus, remains a

Station Bus
challenge because of the number of opposite constraints that have

IEC61850 Scope
to be addressed at the same time, such as scalability, reliability,
real-time and cost efficiency. Bay
IEDA IEDB IEDA IEDB IEDA IEDB

Bay Bay
This paper does not provide the way to implement the process Controller Controller Controller

bus, but rather gives means to evaluate and compare different

Process Bus
process bus solutions. In this study, we are performing a quan-
titative evaluation of different process bus architectures along
the reliability, availability, performance and cost criteria. The Breaker
IED
Merging Breaker Merging
Unit IED Unit
performance criteria encompasses the evaluation with respect
to the real time constraints as well as the overall load of the Breaker
IED
Merging
Unit

communication infrastructure. The cost criteria identifies the Circuit Breaker CTs / VTs Circuit Breaker CTs / VTs

initial investment cost for a given solution and also evaluates Circuit Breaker CTs / VTs
the maintenance cost over the life time of the process bus. The
main contributions of this paper are (a) to give a full evaluation
by integrating all relevant criteria; (b) explicitly explain how the
Fig. 1. Illustration of the station and process bus as envisioned by IEC
different evaluations are performed; (c) to integrate all elements
61850.
for the reliability calculation, especially the optical fibres which
are usually omitted but considered as the least reliable elements;
(d) to evaluate the latest redundancy protocols for automation
networks such as PRP and HSR; and (e) to evaluate architectures gateway by transferring IEC 61850 trip commands to primary
that take full advantage of IEC 61850 by deploying protection switchgear and providing position status back to the IEDs. The
and control functions in innovative ways. main challenges when implementing process bus architectures
are to combine (a) the scalability requirement (a throughput
I. I NTRODUCTION of almost 5Mbits/sec is generated by each merging unit1 ); (b)
While smart grid has a different meaning for each actor of the real time requirement (IEC 61850-5 states that each trip
the electrical grid, a common understanding is the ability to signal and sensor value should not take more than 3ms to be
access any data from anywhere. The need for harmonized data transmitted for transmission applications); (c) high reliability
models and communication protocols is therefore obvious and and availability; and (d) cost efficiency. While addressing
has been identified by NIST in 2009 [1]. those requirements individually can already be challenging,
A key element of the electrical grid is the substation and its the combination of all of them into one single solution is
associated automation system. IEC 61850 [2] is the standard particularly difficult because of their opposite implications (cf.
for substation automation defining a common data model as figure 2) . Indeed, a cost efficient solution usually leads to
well as a set of communication protocols. IEC 61850 envisions a less reliable architecture (cheaper and fewer components),
a substation automation system made of two main buses, called while a scalable solution makes it harder to guarantee the real-
station bus and process bus (cf. figure 1). While the station bus time constraints.
has been successfully adopted and deployed by major suppliers In this paper, we are performing a quantitative evaluation of
and integrators around the world, the process bus is taking off different process bus architectures along the reliability, avail-
with several pilot projects; moreover as of end of 2009 no ability, performance and cost criteria. The performance criteria
commercial solution is offered conforming to the standard. encompasses the evaluation with respect to the real time
The process bus is connecting the protection and control constraints as well as the overall load of the communication
devices (IED or Intelligent Electronic Device), merging units infrastructure. The cost criteria identifies the initial investment
and breaker IEDs. A merging unit (MU) acts as a sensor cost for a given solution and also evaluates the maintenance
gateway by sampling the data from current and voltage trans- cost over the life time of the process bus. It is worth noting
formers (CTs/VTs). A breaker IED acts as actuator and sensor that while the performance criteria is clearly specified by the
Corresponding author: [email protected]
ABB Inc., Corporate Research 1 [3] specifies a sending rate of 80 samples per cycle, which implies a
940 Main Campus Dr., Suite 300 frequency of 4KHz for a 50Hz grid, with an approximate message size of
Raleigh, NC 27612, USA 150 bytes

978-1-4244-6551-4/10/$26.00 ©2010 IEEE

 2

On the other hand, the availability of an element i is defined

as
M T T Fi
Ai =
M T T Fi + M T T Ri
with MTTR representing the mean time to repair. The overall
reliability and availability of a system made of n components
in series are defined by
n

Fig. 2. Illustration of the different constraints to be satisfied when imple- Rs (t) = Ri (t)
menting a process bus architecture. i=0

and n

standard, reliability, availability and cost criteria will be driven As = Ai
by customers. In comparison with evaluations found in the i=0
current literature such as [4] (which focuses on reliability, For a system made of n parallel components,
availability and relative investment cost), [5] (focus on relia-
bility and availability), [6] (focus on reliability), [7] (focus on Rs (t) = 1 − (1 − Ri (t))n
investment cost) and [8] (focus on reliability and availability),
and
the contribution of this paper is threefold:
As = 1 − (1 − Ai )n
1) to explicitly present the mathematical equations for
each evaluation in order to show how the performance
figures are calculated. The complexity of such equations, B. Cost
which are missing in the current literature, makes the The cost criteria is made of the initial investment cost and
verification of the evaluation results difficult; maintenance cost for the secondary equipment related to the
2) to broaden the evaluation by integrating all relevant cri- process bus. While the investment cost is straightforward to
teria (i.e. reliability, availability, cost and performance), evaluate by summing the price of each element (only material
as well as all physical elements which are necessary for cost is considered), the maintenance cost is defined as
the reliability calculation (the optical fibres are omitted n

in the current literature but actually are the least reliable M= Pi × λ i
elements); i=0
3) to evaluate the architectures that take full advantage of with Pi and λi are the price and the failure rate of element i
IEC 61850 by deploying protection and control functions respectively.
in innovative ways, as well as the latest redundant
protocols for automation networks such as PRP [9] and
C. Performance
HSR [10].
The purpose of this paper is not to identify the best architecture Two criteria are defined for performance: average network
but rather to evaluate their applicability for both transmission load, L, and worst case transmission delay, D. The average
and distribution substations. network load is considered at the switch level, i.e. the amount
of traffic a given switch has to deal with. The average network
The reminder of this paper is organized as follows. Section load is only taking into account the dominant traffic, i.e.
II explicitly defines the different criteria used to perform the sample values, leaving aside GOOSE, MMS and IEEE 1588
evaluation. Section III presents the different process bus so- traffic. The worst case transmission delay is computed as the
lutions, while section IV performs the quantitative evaluation. delay a trip message needs to be sent from a protection IED to
Finally, before the concluding section, section V compares and a breaker IED. The worst case transmission delay is computed
analyses the different evaluated solutions. under failure condition, i.e. when one element (either an IED,
switch, merging unit or optical fibre) is out of service. It is
II. C RITERIA DEFINITION
worth noting that the transmission delay definition is slightly
A. Reliability and Availability different from the transfer time defined in IEC 61850-5 since
When considering a repairable system, reliability and avail- it does not include the communication stack processing times.
ability are respectively defined as the probability of a system
not failing for a given period of time and as the probability of III. P ROCESS BUS SOLUTIONS
a system working at a given time [11]. Assuming a constant
A. Substation topology
failure rate, i.e. λi (t) = λi , the reliability of an element i is
defined as Five process bus architectures are considered for the evalua-
Ri (t) = e−λi (t) tion, starting from a simple solution to advanced architectures
based on the communication ability among all devices enabled
and is related to the mean time to failure without repair by by the standard. Each process bus solution relates to a com-
 ∞
mon substation topology made of 5 bays. Each bay requires
M T T Fi = Ri (t)dt
0
a dedicated control function and a primary and redundant
 3

protection functions. The primary and redundant protection Bay 1 Bay 2 Bay 3 Bay 4 Bay 5

functions must be located on distinct physical devices, while C P1 P2 C P1 P2 C P1 P2 C P1 P2 C P1 P2

the control function can be either located on a dedicated device

or on a shared device as allowed by IEC 61850. Moreover
each bay contains one current transformer and one voltage S S S S S S S S S S

transformer connected to at least one merging unit; and one

primary switchgear connected to at least one breaker IED. It is
worth noting that the interlocking protection function required MU MU BI BI MU MU BI BI MU MU BI BI MU MU BI BI MU MU BI BI

by the control function is not addressed in this evaluation,

since it is usually implemented at the station bus level.

CT VT CB CT VT CB CT VT CB CT VT CB CT VT CB
B. Process Bus Solution #1
C Control MU Merging Unit CT Current Transformer
The first process bus solution connects each element of one Optical Fiber
P1 Primary Protection BI Breaker IED VT Voltage Transformer
bay to a single switch as described in figure 3. © ABB Group
Copper Wire
November 17, 2009 | Slide 1
P2 Backup Protection S Switch CB Breaker

Bay 1 Bay 2 Bay 3 Bay 4 Bay 5 Fig. 4. Process Bus Solution #2 consisting of a fully redundant communi-
C P1 P2 C P1 P2 C P1 P2 C P1 P2 C P1 P2 cation and process device infrastructure.

Bay 1 Bay 2 Bay 3 Bay 4 Bay 5

S S S S S
C P1 P2 C P1 P2 C P1 P2 C P1 P2 C P1 P2

MU BI MU BI MU BI MU BI MU BI

MU MU BI BI MU MU BI BI MU MU BI BI MU MU BI BI MU MU BI BI
CT VT CB CT VT CB CT VT CB CT VT CB CT VT CB

C Control MU Merging Unit CT Current Transformer

Optical Fiber
P1 Primary Protection BI Breaker IED VT Voltage Transformer
Copper Wire CT VT CB CT VT CB CT VT CB CT VT CB CT VT CB
© ABB Group
November 17, 2009 | Slide 1
P2 Backup Protection S Switch CB Breaker

C Control MU Merging Unit CT Current Transformer

Optical Fiber
Fig. 3. Process Bus Solution #1 consisting of one switch per bay and no P1 Primary Protection BI Breaker IED VT Voltage Transformer
inter-bay communications. © ABB Group
Copper Wire
November 24, 2009 | Slide 1
P2 Backup Protection S Switch CB Breaker

Fig. 5. Process Bus Solution #3 consisting of a ring communication

infrastructure.
C. Process Bus solution #2
The second process bus solution, cf. figure 4, introduces a
redundant communication infrastructure using the PRP proto- E. Process Bus Solution #4
col [9]. Each IED, MU and breaker IED are double attached The fourth process bus solution, c.f. figure 6, is similar to
to two different switches, while the primary equipment is the previous one except that the control function is redundant
connected through two different MUs and breaker IEDs. and is located on the same device as the protection functions.
This architecture eliminates any single point of failure for The behaviour of the redundant control function is not the
the secondary equipment. The PRP protocol is only used focus of this paper, but a classical hot/standby redundancy
between the IEDs, the MUs and the breaker IED. The primary scheme can be applied here (with some careful handling of the
equipment (CT, VT and breaker) is wired via copper wires to select before operate command), while the protection functions
their respective process bus devices. run as hot/hot. Likewise solution #3, two ring protocols are
considered: RSTP and HSR.
D. Process Bus Solution #3
The third process bus solution, c.f. figure 5, uses a ring F. Process Bus Solution #5
communication infrastructure instead of a star as in solution Finally the last process bus solution, cf. figure 7, implements
#1 and #2. Each device incorporates a network switch func- an inter-bay ring communication infrastructure. The inter-bay
tionality and is connected to its two direct neighbours. For communication is achieved through HSR QuadBox devices.
this solution, two ring protocols are considered: the classical The role of an HSR QuadBox is simply to couple different
Rapid Spanning Tree Protocol or RSTP [12] and the High HSR rings. Each bay has one dedicated IED executing the con-
availability Seamless Redundancy protocol or HSR [10]. trol and protection functions, while the redundant protection
 4

Bay 1 Bay 2 Bay 3 Bay 4 Bay 5 of element used in the solutions as well as their relative cost.
C1+P1 C2+P2 C1+P1 C2+P2 C1+P1 C2+P2 C1+P1 C2+P2 C1+P1 C2+P2 To provide a fair comparison with the current literature, the
reliability figures for the IEDs and switches are taken from
[4], the merging units and breaker IEDs ones are from [5], and
the optical fibre one is from [13]2 . The prices listed only give
an order of comparison between elements. We consider each
merging unit sending IEC 61850-9-2 messages conforming to
MU MU BI BI MU MU BI BI MU MU BI BI MU MU BI BI MU MU BI BI
[3], i.e. at 4KHz with an average payload per message of 150
bytes. The network is assumed to be run at a speed of 100
Mbits/sec and the transmission delay introduced by the optical
fibres is negligible.
CT VT CB CT VT CB CT VT CB CT VT CB CT VT CB
Device MTTF MTTR Availability Relative
C Control MU Merging Unit CT Current Transformer (years) (hours) (%) Investment
Optical Fiber
P1 Primary Protection BI Breaker IED VT Voltage Transformer
IED 100 24 99.9973 9
© ABB Group
Copper Wire IEDswitch 90 24 99.9969 10
November 19, 2009 | Slide 1
P2 Backup Protection S Switch CB Breaker Switch 50 24 99.9945 7
MU 150 24 99.9982 8
Fig. 6. Process Bus Solution #4 consisting of a ring communication infras- BI 150 24 99.9982 8
tructure with two IEDs per bay running protection and control redundantly. MUswitch 135 24 99.9979 9
BIswitch 135 24 99.9979 9
HSRRed;Quad-box 135 24 99.9979 5
Optical Fibre 50 24 99.9945 0.1
and control functions are implemented in a device common
TABLE I
to all bays. The common device accesses the HSR rings MTTF, MTTR AND PRICE VALUES FOR EACH ELEMENT USED IN THE
via the HSR RedBox (redundancy box) which implements a PROCESS BUS SOLUTION

gateway functionality between the HSR and non-HSR world

(e.g. between HSR and PRP in this case). The common device
is connected to two different HSR RedBox devices using the From a reliability and availability point of view, the overall
PRP protocol. Finally, two breaker IEDs per bay are attached system is considered working when one control function has
to each breaker, while only one merging unit is present access to its associated breaker, and when at least one protec-
per bay. However, thanks to the inter-bay communication tion function has access to the current and voltage transformer
infrastructure, one merging unit can be out service since the as well as the breaker. Each bay must have a working control
protection devices can re-construct the missing values from and protection function at all time. The following evaluations
the reminding merging units. are achieved for the pure process bus elements only, i.e. the
primary equipment connected to the merging units (CTs, VTs,
Bay 1 Bay 2 Bay 3 Bay 4 Bay 5 breakers and copper wires) are not included in the calculations.
(C2+P2)
{ Bay1 ; Bay2 ; Bay3 ; Bay4 ; Bay5 }

RB C1+P1 C1+P1 C1+P1 C1+P1 C1+P1 RB

A. Process Bus Solution #1
From a reliability and availability point of view, the first
process bus solution has all the elements in series except for
QB QB QB QB the two protection devices of each bays which are in parallel.
Therefore:

Rs1 (t) = RC (t) × (1 − (1 − RP 1 (t)).(1 − RP 2 (t)))
MU BI BI MU BI BI MU BI BI MU BI BI MU BI BI

× RS (t) × RM U (t) × RBI (t)

5
3
CT VT CB CT VT CB CT VT CB CT VT CB CT VT CB × ROF (t)
t
C Control MU Merging Unit CT Current Transformer QB HSR QuadBox
with RC (t) = e− 100
HSR protocol t
and RP 1 (t) = RP 2 (t) = e− 100
P Protection BI Breaker IED VT Voltage Transformer
PRP protocol
© ABB Group
November 24, 2009 | Slide 1
RB HSR RedBox S Switch CB Breaker Copper Wire t
and RS (t) = e− 50
t
Fig. 7. Process Bus Solution #5 with an inter-bay ring communication and RM U (t) = RBI (t) = e− 150
infrastructure and a common redundant and control protection scheme.  ∞
M T T Fs1 = Rs1 (t)dt
0

IV. E VALUATION 1.9 years

To objectively evaluate and compare the different process

bus solutions, a number of parameters have to be agreed upon. 2 While [13] gives the failure rate per kilometre for the optical fibre, we are
Table I summarizes the MTTF and MTTR values for each type assuming a maximum length of less than 100 meters for each connection.
 5

Similarly, the system availability is equal to: C. Process Bus Solution #3


As1 = AC × (1 − (1 − AP 1 ).(1 − AP 2 )) This process bus solution introduces a ring communication
5 infrastructure with the merging units still being redundant, as
× AS × AM U × ABI × A3OF well as the breaker IED. Reliability and availability calcula-
tions for ring topologies are more complex than series and/or
=99.858550% parallel topologies. A common method is to enumerate the
different combinations of devices working and not working
The initial cost of the first solution or investment cost is
that satisfy the reliability definition of the system. Therefore,
equal to Is1  252.5 while the maintenance cost is equal to:
 for each bay presented in figure 5, there are 14 elements
1 1 1 leading to 214 or 16384 different combinations to evaluate.
Ms1 = (10 × × 3) + (5 × )) + (8 × )
100 50 150 However, a good approximation of the calculation can already
1 1 
+ (100 × ) + (0.1 × × 7) × 5 be achieved by only enumerating the number of valid combi-
250 50 nations of all elements working minus one, and for all minus
=2.6/year two elements workings for the least reliable ones only, i.e. the
optical fibres. The contribution of the other combinations to
From a performance point of view, the average network the overall reliability is negligible for the precision required
load seen by the switch is equal to Ls1 = 4000 × 150 × 8 = for this evaluation. Therefore, for this solution the reliability
4.8M bits/sec. The worst case trip delay in normal conditions and availability is equal to:
is simply the time needed by the switch to send the trip, i.e.
150 × 8/100.106 = 12μ sec, however when one element fails, 
e.g. the switch, the transmission delay is theoretically equal Rs3 (t)  RIEDsw (t)(α(t) + 2.β(t) + 2.γ(t)
to 24 hours, i.e. the time required to repair the switch, and 5
therefore considered infinite: Ds1 = ∞ . + 2.δ(t) + 13.(t))
2 2 2
with α(t) =RIEDsw (t) × RM U sw (t) × RIBsw (t)
B. Process Bus Solution #2 7 6
× (ROF (t) + 7.ROF (t).(1 − ROF (t)))
The second process bus solution is based on PRP and
2
doubles the communication infrastructure between the sec- and β(t) =RIEDsw (t).(1 − RIEDsw (t)) × RM U sw (t)
2 7
ondary equipment. For each bay, the control IED needs to × RIBsw (t) × ROF (t)
be working while one of the two protection IED can fail. 2
and γ(t) =RIEDsw (t) × RM U sw (t).(1 − RM U sw (t))
For each protection IED, one connection can fail. The same 2 7
× RIBsw (t) × ROF (t)
pattern applies for the redundant switches with the redundant
2 2
connection towards the merging units. Finally the merging and δ(t) =RIEDsw (t) × RM U sw (t)
7
units and breaker IEDs are also redundant. In this context, × RIBsw (t).(1 − RIBsw (t)) × ROF (t)
the overall reliability and availability is equal to: 2
and (t) =RIEDsw 2
(t) × RM 2
 U sw (t) × RIBsw (t)
Rs2 (t) = RC (t) × (1 − (1 − α(t))2 ) 5
× ROF (t).(1 − ROF (t))2
× (1 − (1 − RP 1 (t)).(1 − RP 2 (t)))
5
× (1 − (1 − RM U (t))2 ) × (1 − (1 − RBI (t))2 ) Therefore,

with α(t) =RS (t) × ROF (t) × (1 − (1 − ROF (t))2 )3 M T T Fs3 3.9 years
and As3 99.984737 %

and,
M T T Fs2 7.0 years The investement cost is equal to Is3 = 333.5, while the
maintenance cost is equal to Ms3 = 3.0/year.
Therefore, the availability is equal to: From a performance point of view, the traffic load seen
by each element of the ring is considered. Therefore, likewise
As2 99.986295%
solution #2, the two merging units produce a throughput of 9.6
Mbits/sec. For the trip transmission delay, two ring protocols
are evaluated: RSTP and HSR. For normal configurations, the
The investment cost is then equal to Is2 = 372.0, while the transmission delay is equal for both protocols, i.e. 12μsec,
maintenance cost is equal to Ms2 = 3.9/year . which corresponds to the transmission delay on the direct link
The average traffic load for each switch is the sum of between P2 and the breaker IED. When one element fails, such
the traffic generated by each merging unit, i.e. Ls2 = as the link between P2 and the merging unit, the transmission
9.6M bits/sec. The transmission delay of a trip is equal to delay becomes DHSRs3 = 3 × 12 = 36μsec for HSR which
12μsec if all elements are working and when one element is the delay between P1 and the first breaker IED after the
fails. merging unit. For RSTP, considering a reconfiguration delay of
 6

100 ms, the transmission delay is equal to DRST P s3 = 100.36 E. Process Bus Solution #5
msec.
This process bus solution introduces a inter-bay ring com-
munication infrastructure. In this case, the reliability calcula-
D. Process Bus Solution #4 tion cannot be individually evaluated for each bay as for the
Compared to the previous process bus solution, the differ- previous evaluations but need to be addressed globally. There-
ence is the introduction of a redundancy for control function fore, the number of combinations is too big to be manually
and the merging of protection and control function onto one written (259 combinations). To have a good approximation of
device. The reliability and availability is evaluated as previ- the reliability, the calculation needs to take into account at least
ously, i.e. by enumerating the valid combinations of elements all the combination for up to all minus five elements failing at
working and not working with up to two elements failing at the same time which brings the number of combinations down
the same time. Therefore, for this solution the reliability and to 242,825. While this can be easily handled by a software
availability is equal to: application, we cannot reproduce in this paper the reliability
equation due to paper length limitation. For this solution, the
 reliability is equal to:
Rs3 (t)  α(t) + 6.β(t) + 2.γ(t)
+ 2.δ(t) + 2.(t) + 9.ε(t) M T T Fs4 5.9 years
5
2.ζ(t) + 2.η(t) + θ(t) As4 99.9863 %
2 2 2
with α(t) =RIEDsw (t).RM U sw (t).RQBsw (t)
6
.ROF (t)
2
and β(t) =RIEDsw 2
(t).RM 2 The investment cost is equal to Is4 = 273.2, while the
U sw (t).RQBsw (t)
5 maintenance cost is equal to Ms4 = 2.3/year.
.ROF (t).(1 − ROF (t))
2
and γ(t) =RIEDsw (t).(1 − RIEDsw (t)).RM U sw (t) The average load for this process bus solution is the sum of
2
.RQBsw 6
(t).ROF (t) the loads generated by each merging unit, with each of them
2 generating two streams of data corresponding to the primary
and δ(t) =RIEDsw (t).RM U sw (t).(1 − RM U sw (t)) equipments of its own bay and the neighbouring one. There-
2 6
.RQBsw (t).ROF (t) fore, Ls4 = 4.8×5 = 24 Mbits/sec. From a delay transmission
2
and (t) =RIEDsw 2
(t).RM U sw (t).RQBsw (t)
point of view, the delay ranges from Ds4 = 12μsec for all
6 bays, when no element fails, to Ds4 = 8 × 12 = 86μsec when
.(1 − RQBsw (t)).ROF (t)
2 2 2
considering the failure of the five main IEDs in the bays.
and ε(t) =RIEDsw (t).RM U sw (t).RQBsw (t)
4
.ROF (t).(1 − ROF (t))2
and ζ(t) =RIEDsw (t).(1 − RIEDsw (t)).RM U sw (t) V. A NALYSIS
2 6
.(1 − RM U sw (t)).RQBsw (t).ROF (t)
This section first summarizes the results of the architecture
2
and η(t) =RIEDsw (t).(1 − RIEDsw (t)).RM U sw (t) evaluations (c.f. table II) performed previously along the
6
.RQBsw (t).(1 − RQBsw (t)).ROF (t) criteria defined in section II and then analyses and compares
and θ(t) 2
=RIEDsw (t).RM U sw (t).(1 − RM U sw (t)) the different architectures, in order to finally classify them with
6
respect to application cases found in utility networks.
.RQBsw (t).(1 − RQBsw (t)).ROF (t)
PB #1 PB #2 PB #3 PB #4 PB #5
Therefore, MTTF 1.9 7.0 3.9 5.1 5.9
(years)
M T T Fs3 5.1 years Availability 99.8585 99.9862 99.9847 99.9862 99.9863
(%)
As3 99.9862 % Relat. Invest. 252.5 372.0 333.5 283.0 273.2
cost
Maint. cost 2.6 3.9 3.0 2.5 2.3
(per year)
The investement cost is equal to: Is3 = 283.0, while the Traffic load 4.8 9.6 9.6 9.6 24
maintenance cost is equal to: Ms3 = 2.5/year (Mbits/sec)
Transm. delay ∞ 12 36 24 96
From a performance point of view, the traffic load seen (μsec)
by each element of the ring is the same as the previous
TABLE II
architecture. From a performance point of view, the delay E VALUATION RESULTS FOR THE DIFFERENT PROCESS BUS SOLUTIONS .
transmission of a trip signal in case of a failure of the right
side IED host the protection and control functions of each bay
is equal to DHSRs4 = 12 × 2 = 24μsec.
 7

Weights Process Bus Process Bus Process Bus Process Bus Process Bus
Transmission Distribution #1 #2 #3 #4 #5
Reliability ++ + -- ++ 0 + +
Availability ++ ++ -- ++ + ++ ++
Relative Investment cost 0 ++ ++ -- 0 0 +
Maintenance cost + + + -- - 0 ++
Network load 0 0 ++ + + + --
Network trip delay ++ 0 -- ++ 0 + -
Transmission Distribution -12 -1 10 0 1 2 7 5 7 9
TABLE III
C OMPARATIVE EVALUATION OF EACH PROCESS BUS SOLUTION BASED ON THEIR APPLICABILITY TO TRANSMISSION AND DISTRIBUTION SUBSTATIONS .
T HE WEIGHTS ARE REPRESENTED BY (++) FOR +2, (+) FOR +1, (0) FOR 0, (-) FOR -1 AND (- -) FOR -2.

A. Process bus analysis Solution #4 leaves the process level unchanged compared
Solutions #1, #2 and #3 model the bay level with a classical to architecture #3, but in turn reduces the amount of bay level
(single) control IED and a main and backup IED scheme devices by one and at the same time adds control redundancy,
as solutions are offered today. Solutions #4 and #5 take the utilizing the process bus in a way that information between the
process bus architectures as basis and from there on also process and the bay level can be routed as needed as compared
optimize the bay level IED schemes, with #5 illustrating one to hard-wired setups. Assuming only device costs (and not
of possible long-term views. accounting for software functionality embedded in the IED
devices), this brings down the initial investment costs to only
While solution #1 provides a classical scheme on the bay 75% of the level of architecture #2, and the yearly maintenance
IED level (control IED as well as main and backup protection costs to 65% of the level of architecture #2. However, avail-
IEDs), the underlying process bus architecture provides single ability of the solution is now on par with solution #2, as well
point of failures both from a communication point of view, as as reliability is well above a ring communication architecture
well as from accessing primary equipment through merging with a classical bay-level control/protection scheme.
units and breaker IEDs. This is also reflected in the message
delay becoming infinite as soon as a communication link to Illustrating a bay level architecture with main con-
a sensor or actuator is broken. Representing the architecture trol/protection per bay and central control/protection for a set
with lowest initial investment costs, the follow-up maintenance of bays, solution #5 shows that the process bus can be imple-
costs of the station can be further optimized by other archi- mented cost efficiently for protection functions which require
tectures, specifically using HSR ring communications. inter-bay communications such as differential protection or
busbar protection with minimal impact on system reliability.
Solution #2 illustrates a fully-redundant process bus solution However, network bandwidth and transmission delay will
with communication laid out in a star topology. The bay-level restrict the size (bays) of such solutions.
IED setup has not been changed compared to solution #1;
in contrast, the process bus layer is laid out fully redundant.
B. Applicability of architectures
Availability and reliability of the architecture is among the
highest of the evaluated architectures; however, this comes Looking at requirements for transmission and distribution
with relative investment costs as well as yearly maintenance stations, we can classify the presented architectures accord-
costs being the highest of all compared solutions. One should ingly. Table III classifies the solutions by weighting each
also bear in mind associated commissioning costs due to criteria by type of substations, i.e. either transmission or
the amount of optical fibres to connect and route inside the distribution substations.
cubicles. As there are always two redundant merging units The analysis of table III leads to a classification of the
connected to one switch, the traffic seen on one switch doubles different process bus solution evaluated in the previous section:
as compared to solution #1.
• For transmission level, solution #2 is the most suitable for
Ring communication instead of star topology is introduced high voltage transmission substation automation where
with solution #3, thus eliminating separate switch devices as reliability and availability are main concerns. Moreover,
those are integrated with the respective bay and process level this solution guarantees the delivery of a trip signal in
devices. Remarkably, reliability of this solution is doubled a timely manner even under the failure of one element,
compared to solution #1 while the respective investment costs and generates a traffic load that is easily manageable. It is
are only increased by about 30%. The maximum transmission worth noting that this solution emphasises the needs for a
delay allows for larger ring sizes; however the practical redundant communication infrastructure and a redundant
ring size is limited through the network traffic generated by ring could have been evaluated too. However, a redundant
each additional process bus device. Compared to solution ring requires to quadruple the number of Ethernet ports
#2, the availability decreases slightly by 0.0015%, while on each device and a special logic to handle this kind of
the investment cost is 10% less than for a fully redundant communication which may lower the device reliability
communication architecture. and increase its cost. However, when optimizing the bay
 8

level side (redundant control and protection), solution distribution and functional redundancy have to be conducted
#4 fulfills availability and reliability requirements at the to fully understand the possibilities and implications of the
same level as solution #2 for transmission stations but at IEC 61850 standard.
lower investment and maintenance costs.
• On the contrary, solution #3 as well as forward-looking R EFERENCES
architecture #4 are more suitable for substations at the [1] Office of the National Coordinator for Smart Grid Interoperability,
distribution level where cost becomes equally important “NIST Framework and Roadmap for Smart Grid Interoperability Stan-
to reliability and availability. The difference between dards Release 1.0 (Draft),” National Institute of Standards and Technol-
ogy, U.S. Department of Commerce, Tech. Rep., 2009.
the two architectures is probably a slower acceptance [2] Communication networks and systems in substations - IEC 61850.
rate for #4 because of its non traditional protection and International Electrotechnical Commission , 2003.
control functions deployment. To further optimize costs, [3] C. Brunner et al., “Implementation Guideline for Digital Interface
to Instrument Transformers using IEC 61850-9-2,” UCA International
the possibility to centralize (backup) protection functions Users Group, Tech. Rep., 2004.
for a set of bays as illustrated for architecture #5 is [4] L. Andersson, K.-P. Brand, C. Brunner, and W. Wimmer, “Reliability
another possibility. investigations for SA communication architectures based on IEC 61850,”
in IEEE Power Tech., August 2005.
[5] M. Kanabar and T. Sidhu, “Reliability and Availability Analysis of IEC
VI. C ONCLUSION AND F UTURE W ORKS 61850 Based Substation Communication Architectures,” in The 2009
IEEE PES General Meeting, Calgary, Canada, July 2009.
Along reliability, availability, cost and performance criteria, [6] V. Skendzic, I. Ender and G. Zweigle, “IEC 61850-9-2 Process Bus
we presented and evaluated different architectures applying the and its Impact on Power System Protection and Control Reliability,”
Schweitzer Engineering Laboratories, Inc., Tech. Rep., 2007.
process bus concepts as defined by IEC 61850. Depending [7] M. Adamiak, B. Kasztenny, J. Mazerdeeuw, D. McGinn S.Hodder,
on the actual application requirements, those criteria will “Considerations for IEC 61850 Process Bus Deployment in Real-World
have different weights and therefore influence the selection Protection and Control Systems: a Business Analysis,” in Cigre, 2008.
[8] U.B. Anombem, H. Li, P. Crossley, R. Zhang and C. McTaggart,
of feasible architectures. “Flexible IEC 61850 Process Bus Architecture Designs to Support
A fully redundant architecture both in terms of communi- Life-Time Maintenance Strategy of Substation Automation Systems,”
cation, as well as process bus devices is required when taking in Cigre, October 2009.
[9] H. Kirrmann, M. Hansson, P. Mri, “IEC 62439 PRP: Bumpless Re-
the classic bay level IED scheme into account. Such a solution covery for Highly Available, Hard Real-Time Industrial Networks,” in
provides the highest availability and reliability, but at the same IEEE International Conference on Emerging Technologies and Factory
time the highest investment costs of all solutions compared. Automation (ETFA), 2007.
[10] H. Kirrmann, O. Kleineberg, K. Weber, H. Weibel, “HSR: Zero recovery
For distribution substations, ring communication utilizing HSR time and low-cost redundancy for Industrial Ethernet (High availability
shows a cost-effective alternative. Seamless Redundancy, IEC 62439-3),” in IEEE International Conference
Leveraging the process bus and focusing on cost, the bay- on Emerging Technologies and Factory Automation (ETFA), 2009.
[11] IEEE 493, “Design of Reliable Industrial and Commercial Power
level IED architecture will see changes in a way that also Systems,” Institute of Electrical and Electronics Engineers, Tech. Rep.,
control functionality is becoming redundant, and at the same 2007.
time integrated with protection functionality on the same [12] IEEE 802.3, “CSMA/CD access method and physical layer specifica-
tions.” Institute of Electrical and Electronics Engineers, Tech. Rep.,
IED devices. Such a solution will increase the availability 2005.
to the same level as an architecture with a star configuration [13] “Military Handbook, Reliability Prediction of Electronic Equipment,”
compared to a classical bay IED scheme but at reduced costs US Department of Defense, Tech. Rep., 1991.
both for investments and maintenance. Both transmission and
distribution applications would be applicable.
Forward-looking, schemes which centralize backup func-
tionality and cover a set of bays simultaneously can be
implemented utilizing process bus architectures. However, the
Jean-Charles Tournier is Principal Scientist at ABB Corporate Research
differences to bay-focused architectures are marginal, and since 2007, focusing on network protocols for substation automation. Jean-
may become relevant for station architectures where either no Charles received a MSc in Computer Science in 2002 from the Grenoble
backup is nowadays available and combine backup function- Institute of Technology (INPG, France); a MSc in Micro-Electronic in 2002
from the Joseph Fourier University (UJF Grenoble, France); and a PhD in
ality together with station HMI and gateway functions on one Computer Science in 2005 from the National Institute of Applied Sciences
multi-purpose device, or for transmission stations with specific (INSA, France).
focus on cost.
While the evaluations performed in this paper give a better
understanding of the trade-off for different process bus so-
lutions, the same type of evaluation needs to be conducted
at the overall level of the substation, including the station Thomas Werner is Global Product Manager at ABB Switzerland for Sub-
bus level of the substation automation, but also the primary station Automation Systems since 2009, focusing on Process Bus Solutions.
equipments from the switchyard. Such global evaluation needs He joined ABB Corporate Research as principal research scientist in 2000,
where his interests were on software architectures for real-time systems in
to encompass further criteria such as the level of physical substation automation. Thomas holds a Diploma in Electrical Engineering
and cyber security, as well as labour costs to produce a more from the University of Stuttgart (Germany).
accurate evaluation. In parallel, further investigations of the
impact of the process bus solutions on bay level function