Hybrid Fuzzy Evaluation Algorithm for Power

System Protection Security Assessment

Georg Janick Meyer Maximilian Dauer
Tobias Lorz, Rene Wehner Rainer Krebs
Johann Jaeger
Friedrich-Alexander Universität Erlangen-Nürnberg Siemens AG
Erlangen, Germany Erlangen, Germany
[email protected] [email protected]

Abstract—System Protection Security Assessment is an impor- correct functionality. However, the review is a complex and
tant task in modern energy grids to ensure system security at all. time-consuming task that is still mainly carried out by experts
The assessment system is particularly challenged by multivariate today. In addition, conventional protection coordination ap-
grid structures caused by volatile renewable infeeds.
This paper presents an innovative strategy to evaluate the
proaches and tools reach their limits of practical applicability.
protection relay coordination of system-wide power grids. A way In [3], [4] the authors introduce a fully automated Protection
to calculate the quality of all protection relays based on realistic Security Assessment (PSA) tool for main and backup protec-
simulation data and independent of the protection method was tion. The system response is simulated, rated and visualized,
sought. The hybrid algorithm consists of two major steps. First, so weak points can be identified. However the information
a systematic analysis with various fault simulations is performed
and the measurement results of all relays of all simulations are
is missing, why the settings lead to the respective results
used as database. Subsequently, the use of fuzzy sets allows to and which settings would be better. Several other methods
express the quality of each relay setting regardless of its type. have been published to find optimal settings locally and
Specifically developed for the use of an optimization algorithm, to coordinate distance (DI) and overcurrent (OC) protection
finally, a new protection coordination is determined for an relays, but no system-wide approach was investigated. [5], [6]
adapted version of the IEEE 9 bus grid. The results are validated,
discussed and the effectiveness of the methodology compared to A method to evaluate protection relay settings based on
conventional setting rules. fuzzification was firstly introduced here [7]. The authors
Index Terms—Fuzzy Evaluation, Relay Settings, Optimization
describe there idea to make a statement on the quality of a
Algorithm, Protection Coordination, Protection Security Assess- protection system without simulations.
ment
II. C HALLENGE
I. BACKGROUND
Power system protection relays are currently the only con-
Protection relays have been developed over 100 years ago trollers in the grid that are allowed to automatically trigger
and are crucial for system security: They limit the impact of a topology change of themselves. Their goal is to keep the
faults and can be seen as the last line of defense for the power whole system stable and to minimize the impact of faults
system. [1] on the transmission system as well as to protect humans and
Distribution grids worldwide are confronted with higher nature. Therefore, protection areas define related sections that
enforcement of power electronics, shorter time constants and are isolated coherently during a fault. The coordination of
an increasing number of prosumers participating in the market. all relays and types in a system-wide grid is a difficult task.
Volatile sources and the increasing use of flexible AC trans- The challenge is to develop an adaptive, automated and clear
mission systems (FACTS) lead to changing load flows and thus evaluation system for protection coordination in multivariate
to different and time-dependent fault situations. Especially, grids. It should function fast and reliably and be designed for
protection relays are disregarded in this times, even though the use of a heuristic optimization algorithm to improve the
incorrect settings of not adapted relays have led to numerous coordination of all relays to the same time.
blackouts in recent years. [2]
A necessary step would be frequent and quick reviews of the
A. Protection Principals & Coordination
protection system coordination after major changes to ensure
In the following the two most common types of protection
relays with backup function are introduced: DI and OC protec-
This research was partly supported by the Federal Ministry of Education
and Research (BMBF) within the framework of the project ”Neue EnergieN- tion. Subsequently, it is explained why protection relays often
etzStruktURen für die Energiewende” ENSURE (FKZ 03SFK1N0) have difficulties in measuring accuracy.

21st Power Systems Computation Conference Porto, Portugal — June 29 – July 3, 2020
PSCC 2020
 1) DI (21): The distance or impedance protection principle
is based on the evaluation of voltage and current and calculates
the resulting fault resistance and reactance XF locally at the
0
point of installation. Based on the known reactance XL of
the surrounding equipment the distance lF from the relay to
the fault location can be determined by formula (1) and (2).
If the fault occurs in the own protection area the relay should
trip immediately. Otherwise it can work as a time delayed
backup for following relays in the surrounded protection
areas. External influences on the measurement distorting
the calculated impedance can lead to a wrong fault location
determination and would thus cause a gross violation of
Fig. 1. Distance protection: Grading chart for a radial network
selectivity. The facts are explained in detail in section II-B.
Therefore, most protective methods use time grading at the
borders of the protective areas. Safety margins are usually command time depends on the amount of short-circuit current
considered of 10 % to 20 % to the end of the next protection (51) or not (50). If the relay uses in addition to the magnitude
area. The resulting first distance zone is to clear faults within the direction of current for the tripping decision, it has the code
its own area with no intentional time delay. The second zone designation 67. Figure 2 shows the tripping characteristic of
is to cover the borders of the protective area securely. Time an independent OC relay with two tripping stages. The high
delays are in the order of 15 to 30 cycles of the fundamental current stage I>> has to trip short-circuits without delay time.
frequency. Following zones are used as backup and should The overcurrent stage I> is used as backup and overload stage.
be coordinated with zone 1 and zone 2 of subsequent relays. Typical setting rules can be seen in formula (4).
To prevent unselective tripping during normal operation the
High voltage grids are designed primarily meshed, so that overload stage must be higher than the highest load current
there are usually several subsequent lines from the protection that can occur. The short-circuit stage depends on the short-
relay’s point of view. For the protection engineer it is difficult circuit level and must be coordinated with subsequent relays.
to coordinate all relays properly. Various protection coordi-
nation philosophies exist and depend on the respective grid I>> = (1.5...4.0) · Ir
situation. The IEEE Guide for Protective Relay Applications (4)
I> = (1.1...1.5) · Ir
to Transmission Lines (C37.113) suggests settings based on
formula (3). [8]
Figure 1 demonstrates the grading for such a complex
structure based on formula (3). Following the time grading
path of relay R1, it is obvious that the zone reach of X3 (R1)
is overlapping X3 (R2). This could cause an over-function.
Whereas, X2 (R4) has no backup protection of X3 (R1) at
the end of line LBE . This circumstance could lead to an
under-function and fault clearing problems. Those are typical
challenges for DI relays that are installed on a line with a
short and long line following. Fig. 2. Overcurrent protection: Grading chart for a stub line
nU o
k
X F = Img (1)
Ik B. Adverse Effects for Measuring Accuracy
XF Especially in the practical use of DI relays, a number of
lF = 0 (2) influencing factors that adulterate the measurement results
XL
must be taken into account. The greatest inaccuracies are
X1 = 0.8 · XLAB caused by phase shift due to long lines, size of the fault
X2 = 1.2 · XLAB (3) impedance, zero sequence mutual coupling, intermidiate
infeed and impedance change through parallel lines. The last
X3 = XLA B + Xmax,next two effects show how strongly the result of the impedance
2) OC (50/51/67): OC relays are the most commonly used measurement can be influenced and are explained in the
protection relay type. They can be used as main and backup following.
protection in transmission and distribution systems. They are
available with various timing characteristics to coordinate with 1) Parallel Lines: in Figure 3 a simple grid model can be
other protection relays and to protect specific equipment. seen. If all three switches are open the relay R1 measures the
The most important difference in function lies whether the correct distance to the fault, which can be calculated based on

21st Power Systems Computation Conference Porto, Portugal — June 29 – July 3, 2020
PSCC 2020
 0 0
formula (5). Under the condition that XL,AB equals XL,BC
the result is a straight line (1-black).
If the switch S2 is closed, the upper and lower line between
node B and C form a parallel line, which reduces the fault
reactance seen by the relay R1 (2-green) and can be calculated
based on formula (6). If the fault is directly at node C (y = 1)
the measured reactance of line LBC is halfed and the relay
just measures XL,AB + 0.5 · XL,BC . Should this not be taken
into account, unselective tripping would be the result.

( 0
x · XL,AB fault on LAB
Xmeas. (R1) = 0 (5)
XL,AB + y · XL,BC fault on LBC
y
Xmeas. (R1) = XL,AB + y · (1 − ) · XL,BC (6)
2
2) Intermediate Infeed: An increasingly common problem
of decentralized renewable energies is the intermediate infeed. Fig. 3. Influence of various effects caused by different switching states of
An example can be seen in Figure 3 (3-red line). Switch S1 S1, S2 and S3 on the measured reactance of relay R1
is closed and S2 and S3 are open. During a fault on line
LBC relay R1 measures just half of the short-circuit current
(IA1 = 0.5 · IA ) supplied by the source connected to node to generate realistic and reliable reference data based on
A. The current IA2 flows from the node B in addition to simulations. Depending on the required attention to detail,
the fault location and thereby causes an additional voltage every X% of every line of the grid a short-circuit fault
drop from node B to the fault, which also increases the simulations is carried out.
voltage at the location of relay R1. Measuring a higher For each individual simulated fault, all measured fault
voltage but not the total short-circuit current leads to a higher impedances of all DI relays and all fault currents of all OC
reactance measurement. In this case relay R1 measures relays are recorded. The measurement result of each relay is
XL,AB + 2.0 · XL,BC , which means that the relay assumes then assigned to the respective protection area (1 to 3) of the
the fault to be much further away than it is. The measuring relay according to the responsibility for the fault. Therefore,
error is greater, the stronger the intermediate infeed is. for every fault simulation is stored, what each relay measures
and to which protection area the result values belong. Figure
The last two cases (blue) show both effects combined. 4 shows the procedure with an example. A short-circuit is
Especially the course of the dark blue line, which is caused simulated on line LDE . The protection relays R5 and R6 are
by the additional infeed from node C, is sophisticated. The connected directly to that line and thus provide the main
maximum measured impedance is on the line LBC and not protection. Their measurement values are assigned to their
at the end as in all other cases. individual ProtArea1. The relays R3 and R7 see the fault in
forward direction and provide the first backup. Their result
The overall coordination of all protection relays and types values are assigned to their ProtArea2. The relays R1 and
including all possible effects that lead to erroneous measure- R9 provide backup protection for R3 and R7 and add the
ments in all possible grid situations is a highly complex measurement results to ProtArea3.
question. That is why there is usually not the one perfect
setting. The main goal of this paper is to provide a robust
solution with minimal complexity, which is able to evaluate
all protection relay settings easily and clearly. This is presented
in the next section.

III. E VALUATION A LGORITHM FOR S ETTING

O PTIMIZATION
A. Underlying Simulation Data
Traditional approaches are using the resistance and reac-
tance of the connected equipment as reference data to calculate Fig. 4. Allocation of individual measurement values of every relay during
the individual settings for each relay (e.g. equation (3)). As the fault to the respective protection area 1, 2 or 3
in section II-B explained, various effects are influencing the
measurement results. In order to take all possible eventualities This process is repeated for all topologies and possible
into account, a systematic fault analysis is done in advance states of the grid. Thus all eventualities and electrotechnical

21st Power Systems Computation Conference Porto, Portugal — June 29 – July 3, 2020
PSCC 2020
phenomena can be taken into account automatically. At the 1
end all results are merged and form the underlying reference

Degree of membership
data for the following evaluation strategy. 0.8
B. Evaluation Criteria
0.6
Two of the most important criteria when assessing power m security
system protection are dependability and security. m dependability
0.4
Dependability is defined as the degree of certainty that
a protection relay or a system of protection relays operates 0.2
correctly. In contrast, the definition of security is that a
protection relay or a system of protection relays does not
0
operate incorrectly. Specifically, this means that dependability 0 2 4 6 8 10
indicates whether a protection relay trips when its necessary. Reactance X in Ω
Security, on the other hand, is the ability to avoid unnecessary
Fig. 5. Qualitative course of the membership functions security and depend-
tripping behaviour in all operating conditions, as well as ability for DI protection relays
short-circuits and faults outside of the relevant zones, and is
therefore closely linked to the notion of selectivity. The edge
between dependability and security of a protection system is to be 0 and good is the point at which it continues to be
usually narrow, especially since both criteria make conflicting 1. The sigmoid function runs between the two points.
demands on the setting values. [9], [10] • The functions can only accept values in the range [0, 1].
• Both function graphs are opposite to each other but
The new evaluation strategy bases on the fuzzification of usually not symmetrical.
the criteria dependability and security using two different • The optimal point of both functions is between good of
membership functions. The principle is the same for any dependability and good of security. The points are marked
protection method and allows a quick and clear assignment with dashed lines in Figure 5.
of the setting values to both criteria. The calculation of the points good and bad for security
and dependability of all settings of all relays bases on the
C. Membership Function previously generated underlying simulation data. In addition,
Precision is crucial in technical systems, although goals, the setting values of the following protection relays with
limits and consequences are not always precisely defined in the same direction of view are used for some points. The
the practical implementation. They move in a certain frame measured impedance up to the following relay is added to the
and are therefore inaccurate or out of focus. By contrast, actual setting value of it. Those points mark a direct violation
organic systems are inherently imprecise. Be it movements of of selectivity. Formula (7) and (8) give an example of the
an animal, the growth direction of a plant or even the human calculation principle.
mind. These systems are powerful as well as highly complex m security:
and, despite or perhaps because of their blurring, superior to X1bad = XP rotArea1, max
the technical systems. In 1965, Prof. Lofti A. Zadeh of Berkley
X1good = 0.7 · XP rotArea1, min
University published his work on fuzzy sets that forms the
basis of today’s fuzzy logic. Fuzzy sets extend the assessment I>>bad = IP rotArea1, min (7)
of membership of elements to a set from bivalent to gradual. I>>good = IP rotArea1, max
Therefore, the membership is described not only by ”0” for
not belonging and ”1” for belonging, but it is any number as
m dependability:
a degree of affiliation possible.[11]
This kind of consideration is now applied to the setting X1bad = 0.7 · XP rotArea1, min
elements X1 , X2 and X3 of DI relays and I>> and I> of X1good = XP rotArea1, max
OC relays. They are assigned to the membership functions I>>bad = IP rotArea1, (8)
max
m dependability and m security. The process is called fuzzi-
fication. This makes it possible to express independently the I>>good = Imax, operation

quality of each setting over a degree of belonging to both

criteria. The graphs of the membership functions are sigmoid For a good rating of dependability a DI relay needs a
functions, which are subject to the following criteria and can high setting value, so it can clear faults far away and work
be seen in Figure 5: as a backup relay. To get sure not to trip unselective, a
• The slope at the inflection point is 0.5. shorter range is advantageous. Tripping at a high current
• Both sigmoid functions are defined by two points: bad is a more secure setup for OC relays, whereas tripping at
and good. Bad is the point at which the graph continues lower current ensures tripping for complex fault scenarios.

21st Power Systems Computation Conference Porto, Portugal — June 29 – July 3, 2020
PSCC 2020
Through the fuzzy sets, it is now possible to formulate A. Test Grid: Adapted IEEE9 Bus Grid
this kind of evaluation mathematically, which is why it can
be perfectly combined with a heuristic optimization algorithm. The power grid model is an extended and adapted version
of the IEEE 9 Bus grid. It is adapted to have more realistic
influences on the protection system through the added parallel
line and changed length of the transmission lines. In addition,
D. Optimization two stub lines are added and the loads distributed. All trans-
mission lines are protected by two DI relays at the beginning
Optimization is the process of finding values for different
and end of each line. Just the stub lines are only protected
variables xi of a system out of the set X in a fraction of
by one OC relay at the beginning of the line. The generators
time, with which the system is best designed. It means that
get protected by an own OC relay. The line data of the test
the evaluation function f of the system is under the conditions
grid is presented in Table I and the transformer data in Table
F = {f 1, f 2, ..., f n} with the determined values in the global
II. Various grid states the protection system should be able to
maximum or minimum. Such evaluation functions are also
handle are also shown in Figure 6 in blue, red and green and
called goal or fitness functions.
explained in Table III.
Each optimization algorithm is based on different basic
techniques and methods for determining the individual
solutions to an optimization problem. The most simple way
would be to calculate all possible solutions and compare
them, which is not feasible for real problems with reasonable
resources.

The fuzzy evaluation strategy is a multi-criteria optimization

problem, since two membership functions (m dependability)
and (m security) for each setting (X1 , X2 , I> , ..) of every
relay must be maximized simultaneously. In addition, the
defining points (bad and good) of some membership functions
are depending on the actual setting values of following relays.
That means a change of the setting values of one relay will
therefore also lead to new membership functions of other
relays, which in turn will lead to a new overall evaluation.
This circumstance makes the optimization problem NP-hard. Fig. 6. Test Case: Modified IEEE9Bus grid model
By combining all membership functions of all settings of all
relays, it is possible to calculate by the mean of all fuzzy
sets one quality index for the whole protection system. This
TABLE I
type of construction thus represents a fitness function to be T EST GRID : L INE DATA . NAMING DEPENDS ON NODES THAT ARE
maximized, the functionality of which is new and well suited CONNECTED .

for use with an optimization algorithm. From To Length l ReactanceX 0 Reactance X

Name
Node Node in km in Ω/km in Ω
Proven to be efficient in the respect of multi-criteria L45 N4 N5 40 0.4 16.0
problems, we suggest to use the EPSODE particle swarm L56-1 N5 N6 30 0.4 12.0
L56-2 N5 N6 30 0.4 12.0
algorithm, which was firstly introduced here [12]. It is L57 N5 N7 10 0.4 4.0
a hybrid, heuristic, optimization algorithm combining the L69 N6 N9 25 0.4 10.0
advantages of EA (Evolutionary Algorithm), DE (Differential L78 N7 N8 25 0.4 10.0
L89 N8 N9 25 0.4 10.0
Evolution) and PSO (Particle Swarm Opimization). L910 N9 N10 20 0.4 8.0
L1011 N10 N11 10 0.4 4.0
In the following case study we show, how the new method-
ology leads to the best compromise between dependability and
security and therefore to an efficient coordination of all relays. TABLE II
T EST GRID : T RANSFORMER DATA
IV. C ASE S TUDY Prim. Vol. Upri Sec. Vol. Usec Power Sn Reac. xd
Name
in V in V in MVA in %
In this section the hybrid fuzzy evaluation strategy is used T1 110 16.5 50.0 8.0
with the EPSODE optimization algorithm to find the best T2 110 18.0 72.0 8.0
T3 110 13.8 120.0 8.0
settings of all protection relays of the power grid model shown
in Figure 6.

21st Power Systems Computation Conference Porto, Portugal — June 29 – July 3, 2020
PSCC 2020
B. Setting Calculation
In order to be able to classify the optimized values later on,
all settings of all relays are calculated based on formulas (3)
and (4). As a database for the formulas the grid data of the
base grid is taken. The result of the calculation is shown in
Table IV for all DI relays and in Table V for the OC relays.
The settings of the three OC relays protecting the generators
are kept the same for the calculation and optimization case
and can be seen in Table VI. Fig. 8. DI08: Minimum and maximum measured impedances for all grid
states according Table III for protection areas 1 to 3
C. Setting Optimization
The test grid contains of 14 DI and 2 OC relays. That makes
46 fuzzy sets as shown in Figure 5, which need to be optimized
simultaneously by the used EPSODE algorithm.
TABLE IV
T EST C ASE : C ALCULATED SETTING VALUES FOR ALL DI RELAYS BASED
TABLE III ON FORMULA (3) ON THE LEFT SIDE AND OPTIMIZED VALUES ON THE
T EST GRID : VARIOUS STATES OF THE IEEE 9 BUS BASE GRID TO BE RIGHT SIDE IN Ω
HANDLED BY THE PROTECTION SYSTEM
X1 X2 X3 X1 X2 X3
Pumped storage Parallel Line Loads DI-Relay 10 ms 250 ms 500 ms 10 ms 250 ms 500 ms
Name
power plant G1 (blue) L56 (red) (green) calculated optimized
Base grid generator operation in operation 100 % DI01 12.8 19.2 28.0 13.5 18.9 29.8
Case 1 pump operation in operation 100 % DI02 12.8 19.2 25.6 13.5 20.9 30.5
Case 2 generator operation single line 100 % DI03 9.6 14.4 24.0 10.1 20.2 171.1
Case 3 generator operation in operation 50 % DI04 9.6 14.4 28.0 10.0 15.5 56.7
DI05 9.6 14.4 24.0 10.1 13.6 171.2
DI06 9.6 14.4 28.0 10.1 15.5 56.8
As underlying reference data for the evaluation all grid DI07 3.2 4.8 14.0 3.4 9.0 15.6
states as described in Table III are used. On every 10 % of each DI08 3.2 4.8 20.0 3.4 11.4 36.2
line a 3-phase fault is simulated dynamically with the network DI09 8.0 12.0 20.0 8.5 18.6 30.7
DI10 8.0 12.0 14.0 8.4 11.6 19.1
planing tool, PSS©Netomac, and the measured values of all DI11 8.0 12.0 20.0 8.4 13.6 28.6
relays assigned to the corresponding protection areas. Figure DI12 8.0 12.0 29.4 8.5 18.6 39.6
7 shows the minimum and maximum measured impedances DI13 8.0 12.0 20.0 8.4 13.6 25.4
DI14 8.0 12.0 22.0 8.5 20.9 160.1
of protection relay DI08 for all three protection areas of the
base grid. Figure 8 shows those aspects for all grid states
(base grid + case 1 to 3) combined. It is easy to see, that the
measured fault impedances of protection area two and three
are extensive and in both cases strongly overlapping. For all TABLE V
grid states combined it is even worse. That means that both T EST C ASE : C ALCULATED SETTING VALUES FOR THE OC RELAYS OF THE
protection areas make different demands on the setting values STUB LINES BASED ON FORMULA (4) ON THE LEFT SIDE AND OPTIMIZED
VALUES ON THE RIGHT SIDE IN Ir
in terms of security and dependability. The resulting setting
values can be seen in Table IV and V. I>> I> I>> I>
OC-Relay 10 ms 250 ms 10 ms 250 ms
calculated optimized
OC01 4.0 1.2 4.65 1.15
OC02 4.0 1.1 3.12 1.05

TABLE VI
Fig. 7. DI08: Minimum and maximum measured impedances in the base grid T EST C ASE : U SED SETTING VALUES FOR THE OC PROTECTION RELAYS
according Table III for protection areas 1 to 3 OF THE GENERATORS IN Ir

I>> I>
OC-Relay
D. Result 10 ms 2s

To validate and compare both protection setting setups, OCG1

1.1 1.1
OCG2 forward direction both directions
the PSA methodology within the system analysis software, OCG2
PSS©Sincal, is used. It simulates a ”running” fault on all
lines and evaluates the protection relay behavior in terms of

21st Power Systems Computation Conference Porto, Portugal — June 29 – July 3, 2020
PSCC 2020
selectivity and fault clearing time. It also considers the failure
of the main protection to analyze the tripping behavior of
the backup relays. The mechanical delay time of all circuit
breaker is 40 ms. Because the calculated values base on the
base grid, it was taken to be evaluated with both setups. The
results can be seen in Figure 9 for the calculated values and
in Figure 10 for the optimized values. For all other cases the
calculated values achieved similar or worse results, whereas
the optimized values were just as good or even better. The
ranking and meaning of the colours is as following:
• Selective (green): Protection system acted correctly.
• Over-function (yellow): In addition to the responsible
relays, at least one additional relay tripped.
• Under-function (orange): Some of the responsible relays
did not trip, but the fault was cleared by other relays.
• No fault clearing (red): Fault was not tripped till the
end of the simulation.
Using the calculated values, the PSA analysis shows that
in 16 out of 32 investigated situations the protection system
does not work as intended. In a large number of cases, faults
are tripped with under- or over-function and the fault clearing
times are relatively high. The result clearly shows the negative
effects on the measurement due to the intermediate infeed and
the parallel line effects.
For the optimized values, the PSA analysis finds 3
situations for improvement and all concern just the backup
protection. On some sections of the relevant lines the fault
is cleared with an over-function (zone 3). This is because Fig. 9. Test case: Result of PSA analysis for the calculated setting values
all protection relays are using the same delay time for all
tripping zones. A minimal adjustment of the delay time to
improve the protection coordination would solve the problem.
In the future, the delay time could be added to the evaluation
algorithm as a further variable to be optimized. Last but not
least, it is noticeable that the fault clearing times are on
average significantly shorter.

The new hybrid evaluation strategy combining fuzzy sets

with a new way to assign simulation values to protection areas
in cooperation with a search/optimization algorithm proves to
be flexible applicable and produces reliable setting values.
V. S UMMARY
We presented an innovative strategy that is able to find new
protection relay settings for DI and OC protecion relays to the
same time. It is able to find the best compromise in complex
grid structures in terms of security and dependability. It can be
applied for any grid size and considers various grid variations
to act correctly even in unusual and emergency situations.
Todays power grids need new methods for assessing and
optimizing their protection systems. A promising approach
should be generic, flexible and easy to reconstruct. We pro-
posed a modular concept of grid models, a fuzzy evaluation
strategy and optimization algorithm.
We compared protection system settings calculated by ad-
justment rules based on equipment data with a new way to Fig. 10. Test case: Result of PSA analysis for the optimized setting values
use simulation data to compensate for any negative effects on

21st Power Systems Computation Conference Porto, Portugal — June 29 – July 3, 2020
PSCC 2020
the measurement accuracy. Especially in future and rapidly
changing grids, this will provide an advantage.
R EFERENCES
[1] R. Rifaat, “Power System Protective Relays: Principal & Practices,”
Calgary, Canada, November 2016.
[2] R. Krebs, E. Lerch, O. Ruhlem, S. Gal, and Lazar, F. and Paunescu,
D., “Vision 2020: Blackout Prevention by Combined Protection and
Network Security Assessment,” IEEE Power and Energy Society (PES)
- General Meeting, Pittsburgh, PA, USA, July, 2008.
[3] R. Ganjavi, M. Mangold, M. Friedrich, R. Krebs, M. Dauer, and
M. Worch, “Automated Protection Security Assessment of Main and
Backup Protection for Distribution Networks,” International Conference
on Smart Energy Systems and Technologies, Sevilla, Spain, Oktober,
2018.
[4] T. Bopp, R. Ganjavi, R. Krebs, B. Ntsin, M. Dauer, and J. Jaeger,
“Improving Grid Reliability through Application of Protection Security
Assessment,” 12th IET International Conference on Developments in
Power System Protection, Liverpool, UK, March, 2014.
[5] T. S. Sidhu, D. S. Baltazar, R. M. Palomino, and M. S. Sachdev, “A new
approach for calculating zone-2 setting of distance relays and its use in
an adaptive protection system,” Power Delivery, IEEE Transactions on,
vol. 19, no. 1, pp. 70–77, 2004.
[6] Y. Damchi, J. Sadeh, and H. R. Mashhadi, “Optimal coordination of
distance and overcurrent relays considering a non-standard tripping
characteristic for distance relays,” IET Generation, Transmission &
Distribution, pp. 1448–1457, 2016.
[7] M. Dauer, J. Jaeger, T. Bopp, and R. Krebs, “Protection Security
Assessment - System Evaluation Based on Fuzzification of Protection
Settings,” 4th IEEE PES Innovative Smart Grid Technologies Europe,
Copenhagen, Denmark, October, 2013.
[8] “IEEE Guide for Protective Relay Applications to Transmission Lines:
IEEE Std C37.113-2015,” Piscataway, NJ, USA, December, 2015.
[9] G. Ziegler, Numerical Distance Protection: Principals and Application.
Munich: Publicis MCD, 1999.
[10] J. L. Blackburn and T. J. Domin, Protective Relaying - Principles and
Applications. CRC Press, 2006, vol. 3.
[11] Lofti A. Zadeh, Toward a Theory of Fuzzy Systems. Berkley: In:
Electronic Research Labratories, ERL Report No. 69-2, 1969.
[12] M. Dauer, G. J. Meyer, J. Jaeger, T. Bopp, and R. Krebs, “EPSODE Al-
gorithm for System-Wide Protection Coordination,” 19th Power System
Computation Conference, Genua, Italy, June, 2016.

21st Power Systems Computation Conference Porto, Portugal — June 29 – July 3, 2020
PSCC 2020