IEEE Power and Energy Technology Systems Journal

Received 8 December 2015; revised 8 February 2016; accepted 22 March 2016.

Date of publication 4 May 2016; date of current version 22 June 2016.
Digital Object Identifier 10.1109/JPETS.2016.2546282

A Systematic Approach to n-1-1 Analysis

for Power System Security Assessment
PARAG MITRA1 (Student Member, IEEE), VIJAY VITTAL1 (Fellow, IEEE),
BRIAN KEEL2 (Senior Member, IEEE), AND JENI MISTRY2
1 Arizona State University, Tempe, AZ 85287 USA
2 Salt River Project, Phoenix, AZ 80062 USA

CORRESPONDING AUTHOR: V. VITTAL ([email protected])

ABSTRACT Cascading outages can have a catastrophic impact on power systems. One such recent incident
was the September 8, 2011 blackout that affected San Diego and large parts of the southwestern United States.
To prepare for such events, the NERC transmission planning standards require utilities to plan for n-1-1
outages. However, such analyses can be computationally burdensome for any realistic power system owing
to the staggering number of possible n-1-1 contingencies. This paper proposes a systematic approach to
analyze n-1-1 contingencies in a computationally tractable manner for power system security assessment.
The proposed approach addresses both static and dynamic security assessment. The proposed methods have
been tested on the Western Electricity Coordinating Council (WECC) system. The test results show that
a substantial reduction in computational effort is achieved by implementing these methods for an n-1-1
contingency analysis. In addition, the reliability of the proposed methods is evaluated by an exhaustive
contingency analysis.

INDEX TERMS Contingency classification, contingency screening, contingency ranking, dynamic security
assessment, multiple outages, n-1-1 contingency, power system stability, static security assessment.

I. INTRODUCTION Ensuring the safe and reliable operation of a power sys-

C ASCADING outages can have a catastrophic impact

on power system operation and planning. Accordingly,
the North American Electric Reliability Corporation (NERC)
tem requires assessing both the static and the dynamic
security for every n-1-1 contingency [4], [5]. Although
such assessments improve the overall reliability of a
transmission planning (TPL) standards requires utilities to system, it comes with an additional burden of eval-
plan for n-1-1 outages [1]. To comply with the NERC TPL uating a vast number of outages. The number of
standards, utilities must ensure that the system is able to n-1-1 contingencies can be overwhelming, even for
maintain stability and operate within acceptable limits fol- a small power system. For a system with n ele-
lowing such outages. An n-k contingency analysis involves ments, the number of possible n-1-1 contingencies is
studying the impact of simultaneous outage of k compo- given by
nents on the system voltages and line flows [2], [3]. For an
n-1-1 contingency analysis, the two contingencies are applied number of n-1-1 contingencies = n(n − 1). (1)
sequentially rather than simultaneously. System adjustments
if needed are made after the first outage, prior to applying the Due to the large number of possible n-1-1 outages, a sys-
second outage [2]. Due to the possible system adjustments tematic approach to screen n-1-1 contingencies is needed.
that can be made, the order in which the n-1-1 contingen- It is important to identify the contingencies that are criti-
cies occur can be consequential. System adjustments can cal from a system security viewpoint in a computationally
include [2]: efficient manner. The critical contingencies can be conse-
1. change in VAr output of generators; quential, and need to be identified and analyzed prior to the
2. adjustments in the settings of tap changer (ULTC), non-critical cases. To achieve the stated objective this paper
switchable voltage devices (SVD) and phase adjusting proposes:
transformers (PAR); and 1. a method to screen and rank n-1-1 contingencies for
3. corrective measures like remedial action static security assessment (SSA), and
schemes (RAS), transmission switching and special 2. a method to screen and classify n-1-1 contingencies for
protection schemes (SPS). dynamic security assessment (DSA).
2332-7707
2016 IEEE. Translations and content mining are permitted for academic research only.
Personal use is also permitted, but republication/redistribution requires IEEE permission.
VOLUME 3, NO. 2, JUNE 2016 See http://www.ieee.org/publications_standards/publications/rights/index.html for more information. 71
 IEEE Power and Energy Technology Systems Journal

The goal of the paper is to develop screening methods, The proposed techniques are used to perform contingency
which are convenient to implement in practice, at the same analysis for a utility in the WECC system. The following sec-
time being highly accurate in detecting critical contingencies. tions present detailed descriptions of the proposed methods
The proposed methods are intended to be used during the and the results of the study.
operations planning phase. In the following section, some of
the existing contingency analysis methods are discussed and III. CONTINGENCY SCREENING AND RANKING
the need for a new approach is highlighted. FOR SSA
AC power flow solutions capture the effect of an outage on
II. REVIEW OF CONTINGENCY SCREENING METHODS system voltages and line flows better than dc power flows.
For n-1 SSA, performance indices (PIs) based methods have This is particularly significant while analyzing n-1-1 outages.
been widely adopted [6]–[8]. These methods use a first Therefore, an ac power flow based method is proposed here
order estimate of the post-contingency PI value, to mea- to develop a reliable n-1-1 contingency screening and rank-
sure the severity of a contingency. Contingency screening ing method. Reference [22] describes a method for multiple
methods, which use partial or approximate network solu- element contingency screening using line outage distribution
tions to identify voltage and flow violations have also been factors (LODF). The method proposed here uses a similar
proposed [9]–[11]. However, these methods can grossly idea. However, instead of LODFs, it relies on the correspond-
underestimate the severity of a contingency [12]. This is ing ac quantities for the first n-1 outages to screen critical
due to the approximations involved and the highly non- n-1-1 outages.
linear impact of the contingencies on the system voltages
and line flows [12]. In recent times, with the advancement A. CONTINGENCY SCREENING FOR n-1-1 ANALYSIS
in computational capability, performing a complete ac power In an interconnected power system, the outage of a
flow for large systems has become computationally feasible. single component only affects a few branch flows and gen-
Analyzing all contingencies, using ac power flow solutions erator outputs. The majority of branch flows and generator
is far more reliable. The availability of parallel computing outputs throughout the system remain practically unchanged.
infrastructures further eases this process. Hence, many com- Substantial changes in branch flows or generator outputs
mercially available power system analysis packages use ac following an outage indicate that the affected branches and/or
power flow solutions as a tool to perform n-1 contingency generators compensate for the disconnected component.
analysis [13]. Therefore, to account for critical n-1-1 contingencies, can-
For n-1 DSA, it is important to identify contingencies didate branches or generators for the second outage can
that may jeopardize the stability of the system before con- be screened by analyzing their pre-contingent and post-
ducting a complete time-domain simulation (TDS) analysis. contingent state (flows for branches and outputs for gen-
Various contingency screening methods for DSA have been erators) following the first outage. To generate a list of
proposed in literature. References [14]–[18] describe con- n-1-1 branch-branch contingencies, the following steps are as
tingency screening methods based on the transient energy follows.
function. Filtering of contingencies using sparse transient 1. Perform a full ac power-flow for all n-1 contingencies,
energy function is described in [16] and [17]. Extensions followed by the system adjustments if any. Store all the
of static security analysis techniques coupled with applica- branch MVA flows.
tions of artificial neural networks have also been explored 2. For each branch outage, compute the percentages of
for contingency screening and ranking [19]. A coherency- the MVA flow in the outaged branch that appears in
based index for contingency screening is described in [20]. all other branches in the system. This percentage is
Composite indices based on coherency, transient energy con- the ac equivalent of the LODF and is referred to as
version and dot products have been proposed for contingency ACLODF in the rest of this paper. As an example, the
screening [21]. Most of these screening methods compute the ACLODF for a hypothetical line A for outage of line B
required indices by performing short duration TDS only for is computed as follows:
a few cycles after the disturbance.  
MVAn−1,AB − MVAbase,A
For n-1-1 SSA, performing an ac power flow for all ACLODFAB = 100 (2)
possible n-1-1 contingencies is computationally prohibitive. MVAbase,B
The discussed n-1 contingency screening methods are inad- where
equate due to their dependence on first order estimates ACLODFAB is the percentage of the MVA flow on
and localized nature of analysis. Similarly, for n-1-1 DSA, line B that shows up on line A.
screening n-1-1 contingencies using the traditional screening MVAn−1,AB is the MVA flow on line A with line B
methods can also be impractical. This is due to the sheer disconnected.
number of short duration TDS that need to be performed. MVAbase,A is the MVA flow on line A in the base case
Therefore, there is a need to develop methods, which can (all lines in service).
perform n-1-1 contingency analysis in a computationally MVAbase,B is the MVA flow on line B in the base case
tractable manner. (all lines in service).

72 VOLUME 3, NO. 2, JUNE 2016

Mitra et al.: Systematic Approach to n-1-1 Analysis for Power System Security Assessment

3. All branches that have an ACLODF more than a set beyond its emergency rating, the sequential combination of
threshold for an n-1 contingency are screened as can- the two n-1 outages can also be critical. Steps 3 to 6 of
didate branches for the second outage following that the algorithm screen n-1-1 branch-branch outages based on
particular n-1 contingency. these ideas. Step 3 was found to be particularly significant in
4. For every candidate branch screened in step 3, the identifying n-1-1 contingencies that resulted in overloading
MVA flow on that branch due to a combination of any of transformers and lines, which were operating near their
two n-1 outage is estimated. The change in flow is emergency ratings in the base case. Similar justifications
linearly approximated by using the already computed can be provided for generator-branch and generator-generator
ACLODFs. contingency screening process. Although not all screened
5. The n-1-1 contingency list is updated to include all contingencies may be harmful, the screening process sub-
pairs of the n-1 outage and the corresponding screened stantially reduces the total number of n-1-1 contingencies
second outages in step 3. Steps 2 to 4 are performed for that needs to be evaluated in details (by ac power flow).
all n-1 contingencies. The remaining non-critical contingencies, if needed, can be
6. The contingency list is updated to include any com- evaluated by faster approximate methods like a dc power
bination of two n-1 contingencies, which may cause flow.
the estimated flow on a screened branch computed in
step 5, to be above a set percentage of the emergency B. RANKING OF SCREENED n-1-1 CONTINGENCIES
rating of the branch. After the screened n-1-1 contingency list is prepared, the
7. Finally, all duplicate n-1-1 contingencies are removed screened contingencies are analyzed by performing ac power
from the list and the final list of n-1-1 contingency is flows. Based on the power flow results, the contingencies can
formed. be ranked according to their impact on the system. Ranking
Steps 1 to 7 described here are used to generate the contingencies based on their severity enables a planner or
n-1-1 contingency list for branch-branch contingencies. an operator to identify the most critical elements in a power
An identical process is used to generate the n-1-1 contin- system. Additional measures can then be taken for hardening
gency list for generator-branch and generator-generator con- such elements to prevent such outages. Among the various
tingencies. Instead of the ACLODF, the change in the branch indices proposed in the literature [6]–[8], the voltage-based
MVA flows can be used as a criterion to screen generator- index PIV and the flow-based index PIMW as defined [6] are
branch contingencies. Similarly, the change in the reactive used in this paper. The indices are as follows:
power output of generators can be used as a criterion to screen  sp  !2α
nb
generator-generator contingencies. X wbi |Vi | − Vi 
PIV =
Step 3 to step 6 is the core of the proposed method and
i=1
2α 1Vilim
is elaborated here. The ACLODF captures the MVA flow !2α
nl
changes in the system in comparison to the MVA flow on X wli Si
PIMW = . (3)
the disconnected line. This helps to identify the branches that 2α Silim
i=1
compensate for the disconnected branch, and hence important
for system operation after the first outage. Lines with higher The weighting factors wbi and wli can be chosen by
ACLODF have a higher influence on the connectivity of the an operator or a planner based on experience, to assign
transmission system and indicate the lower impedance paths higher weights to critical buses or lines in the system. The
to the loads. For branch-branch outages, if the outage of a parameter α is set to 1 as given in [6].
single branch causes the ACLODF for another branch to be
significantly high, it indicates that this second branch is an C. THE n-1-1 CONTINGENCY SCREENING AND THE
important supply route (lower impedance path) to the same RANKING PROCESS
load area. An outage of a branch with a high ACLODF can The screening and ranking process described in the preced-
lead to considerable weakening of the transmission path to ing subsections are combined to perform the n-1-1 analysis.
the load area. The alternate transmission paths to the load, Fig. 1 provides a flow chart of the complete n-1-1 contingency
which have a lower ACLODF, typically have substantially analysis method.
higher impedance. This can lead to large voltage drops and
possible reactive power problems at the load buses. These D. n-1-1 CONTINGENCY ANALYSIS RESULTS FOR STATIC
lines may also be overloaded due to the excess power being SECURITY ASSESSMENT
routed through them. Close to large generation sources, dis- For testing the n-1-1 contingency screening and ranking
connecting a branch with a high ACLODF can cause signif- method, a power flow case representing the WECC system
icant reduction in the short circuit ratio, leading to voltage was used. Contingency screening and ranking was performed
regulation problems. Hence, the outage of such a branch may for a selected utility’s service area within the WECC sys-
be critical following the corresponding n-1 branch outage. tem. The power flow case used is an actual representation
Furthermore, if two separate n-1 branch outages cause the of a large interconnected power system with 20605 buses,
estimated MVA flow on a branch or branches to increase 17056 lines, 7836 transformers and 2737 generators.

VOLUME 3, NO. 2, JUNE 2016 73

 IEEE Power and Energy Technology Systems Journal

TABLE 2. Criteria for screening n-1-1 contingencies.

FIGURE 1. Flowchart for the contingency screening and ranking

process.

For n-1 and n-1-1 contingency analysis, elements at all

voltage levels above 100 kV were considered. Including
TABLE 3. Results of the screening process.
all transmission lines, transformers and generators within
the selected utility’s service area, the n-1 contingency list
comprises of 403 outage events. For the n-1-1 analysis,
all remedial actions after the n-1 contingencies were limited
to adjustments in voltage control devices (SVD and ULTC),
power control devices (phase adjusting transformers) and VAr
output of generators. The remedial actions were obtained
from the utility supplied data. It was observed that for such
remedial actions, which did not change the system topology
substantially, the order in which the two contingencies were examined in detail, was achieved. As a recommendation,
processed did not have any significant effect on the results. moderate screening thresholds should be chosen to begin with
Therefore, the total number of n-1-1 contingencies is 81003, and these thresholds should be tightened/tuned over time to
ignoring the duplicate contingencies. In this sense, the pro- achieve better efficiency. ac power flows were performed
posed method could be extended to screen n-2 contingencies for the 5549 screened cases to identify the contingencies
as well. However, for n-2 contingencies the ACLODFs are that resulted in violations. The power flow study indicated
computed without performing any system adjustments. that 3115 contingencies resulted in violations of voltage or
Table 1 lists the voltage and flow thresholds used for n-1 flow or both thresholds. To check whether the contingency
and n-1-1 contingency analysis. Any contingency violating screening process is reliable, an exhaustive ac power flow
based study was conducted for the 81003 contingencies.
TABLE 1. Voltage and flow thresholds. The exhaustive study identified the same 3115 contingencies
reported in Table 3 as critical. It should be noted that the
screening method captures a considerable number of non-
critical contingences due to the moderate screening thresh-
olds used in this work. The number of screened cases could
be further reduced by tightening the screening thresholds
mentioned in Table 2. Although, using tight thresholds could
these thresholds is considered as critical. GE-PSLF [14] was result in possible omission of some critical contingencies.
used to perform the power flow calculations. From the power However, even moderate screening thresholds reduce the total
flow analysis of the n-1 contingencies, 6 contingencies were number of contingencies to be analyzed to a manageable
found to violate the thresholds listed in Table 1. Table 2 lists level. Table 3 also lists the number contingencies that lead
the criteria used for screening the n-1-1 contingencies. to only voltage or only flow violations. It can be seen that for
Table 3 presents the results of the screening process. From 38 contingencies no flow violations were observed although
Table 3, it is observed that the screening process substan- these outages led to severe voltage violations in parts of the
tially reduces the number of cases that need to be analyzed system. The voltage violations were due to large voltage
in detail. It is worthwhile to mention that even with the drops on lines, which served as the alternate transmission
moderate screening thresholds (see Table 2) used in this paths to the load area. However, the flows on these lines were
study, a significant reduction in the number of cases to be well below their emergency ratings. This further illustrates

74 VOLUME 3, NO. 2, JUNE 2016

Mitra et al.: Systematic Approach to n-1-1 Analysis for Power System Security Assessment

that contingency screening based on ACLODF is able to As an example, consider the outage of lines 1159-1232
identify cases where voltage violations may be an issue rather and 1217-1133. Fig. 2 shows a schematic one-line diagram
than flow violations. dc power flow based screening methods of the area of interest. The disconnected branches 1159-1232
typically focus on detecting possible flow violations in the
system.
Once the n-1-1 contingencies are screened and ac power
flows are conducted for the screened cases, the bus voltages
and the line flows are recorded. The contingencies are then
ranked based on the indices given by (3) and (4) in [6].
Table 4 lists the choice of weights used for ranking the
contingencies. As mentioned earlier a different weighting

TABLE 4. Weights used for contingency ranking process.

FIGURE 2. Schematic one-line diagram of the load area supplied

by lines 1159-1232 and 1217-1133.

scheme can be used based on the operator/planner’s experi- and 1217-1133 are marked with cross. When these two lines
ence. Table 5 lists the first three contingencies based on are disconnected, the loads in the region are supplied through
voltage based ranking and flow based ranking respectively. 1143-1218-1121/1122 (marked with dotted line), which has a
considerably higher impedance. The large voltage drop on the
TABLE 5. Voltage and flow based ranking of n-1-1 contingencies. higher impedance path and lack of reactive power resources
in the region leads to depressed voltages at the load buses.
Fig. 3 shows the voltage at selected load buses in the region
when the branches are disconnected. The power flow fails

Ranking the contingencies enables an operator or a planner

to identify the most severe outages in the system. Based on
the information whether voltage violations or flow limit vio-
lation are of concern for these contingencies, corresponding
measures can be taken to mitigate such violations.
FIGURE 3. Voltages at selected load buses in the area supplied
E. ANALYSIS OF n-1-1 CONTINGENCIES WHERE by lines 1159-1232 and 1217-1133.
POWER FLOW FAILS TO SOLVE
Since n-1-1 contingencies result in loss of two system compo- to solve for this case due of the abnormally low voltages in
nents, such contingencies may lead to cases where the post- this area, resulting from the outage. Contingencies, that cause
contingent power flow fails to solve. For the WECC case used divergence in power flow solutions, lead to very high or very
here, the power flow fails to solve for 17 n-1-1 contingency low voltages in parts of the system. Hence, these cases are
cases. These contingencies can be examined by perform- critical from a static security perspective.
ing a dynamic study simulating the opening of the second
line. Such a time domain simulation can be performed as IV. n-1-1 CONTINGENCY SCREENING AND
follows. CLASSIFICATION FOR DSA
1. A power-flow is solved for the first outage. The power- The contingency events referred to in this section are
flow solution is taken as the initial condition for the three-phase faults on lines followed by removal of the faulted
time domain simulation. lines from service by appropriate protection devices. Faults
2. A time domain simulation is run and the second branch are large disturbances on the system and have significant
is opened at t=1 sec. impact on transient stability. DSA for n-1-1 contingencies

VOLUME 3, NO. 2, JUNE 2016 75

 IEEE Power and Energy Technology Systems Journal

is done in two steps. The first step is to simulate the initial a few branches compensate for the lost flow on the discon-
contingency and check whether it is stable. If the system is nected branch. This implies that the branch may be closer
stable and reaches an acceptable equilibrium after all auto- to a generation source or a load center where the network
matic controls have acted, the second contingency is simu- is vulnerable. A fault on a line with a high MWLODF can
lated. If the second contingency is stable and an acceptable be critical as the disturbance may be close to a generator or
steady state is reached, the n-1-1 contingency is labeled as group of generators. Subsequent disconnection of the line to
secure. Otherwise, the n-1-1 contingency is labeled as inse- clear the fault can result in a weak post-disturbance network
cure. As discussed earlier, one of the main challenges for due to loss of a main transmission path. A lower value of the
n-1-1 contingency screening and classification is the large MWLODF indicates that only a small fraction of the lost flow
number of cases that need to be evaluated. As such, perform- on the disconnected line shows up in other lines. Faults at
ing simplified fast calculations can become computationally such locations followed by removal of the line do not weaken
prohibitive for realistic power systems. To alleviate the com- the transmission system significantly. It is worthwhile to
putational burden associated with n-1-1 contingency analysis, mention that a higher value of the MWLODF does imply that
a two-stage approach is proposed here. In stage I, a subset a fault on the corresponding line will be more severe than
of n-1-1 contingencies is screened for further classification. on others. However, it is a good measure to identify critical
The screening process is based on the power flow results lines where faults could be possibly severe. The screening
of the n-1 analysis. In stage II, the screened contingencies process is conservative in the sense that not all screened cases
are classified based on kinetic energy gained due to a fault are unstable. It should be noted that stability assessment is
and the change in Thévenin’s impedance (Zth ) at the point not an issue here. The screening and classification process
of interconnection (POI) of the affected generators in the aims to prioritize the analysis of cases, which are critical from
post-fault network. Once the contingencies are screened and a stability assessment viewpoint. The n-1-1 contingencies
classified accordingly, detailed TDSs are conducted to check eliminated by the screening process are categorized as least
whether the contingencies may lead to dynamic security critical (LCR) for stability assessment.
issues.
B. STAGE II: CLASSIFICATION OF SCREENED
A. STAGE I: SCREENING OF n-1-1 CONTINGENCIES
n-1-1 CONTINGENCIES
The screening process is based on the power flow results
Stage II deals with classifying the screened n-1-1 contingen-
of n-1 contingency analysis. The power flow solution after
cies obtained from stage I. The screened contingencies are
a contingency describes the post-disturbance equilibrium of
classified based on two factors:
the system, once all fast control actions have taken place.
The impact of the second contingency on the system is 1. total kinetic energy gained by the machines due to the
dependent on this equilibrium point and hence the n-1 power applied fault, and
flow solution is the starting point for the n-1-1 screening 2. the maximum change in Zth seen at the POI of
process. It should be noted that only stable n-1 contingencies a generator or a group of generators affected by
are considered as the contingencies for the n-1-1 screening. the fault due to removal of the faulted line from
This is a valid consideration as it is meaningless to analyze a service
successive contingency if the initial contingency is not stable. It is assumed that a fault applied on the system will be
The screening process is similar to the screening process pro- cleared in 5 cycles, which is a standard clearing time for
posed for SSA in section III A. However, instead of ACLODF modern circuit breakers rated for 100kV and above [23].
a new quantity MWLODF is used. The MWLODF is the This clearing time can be suitably adjusted for specific
percentage of MW flow in the outaged branch that shows cases without loss of generality. Based on the two measure-
up in all other branches in the system. As an example, the ments the screened contingencies are categorized as criti-
MWLODF in a hypothetical line A for outage of line B is cal (CR), possibly critical (PCR) and non-critical (NCR).
computed as follows: The following steps are followed to classify the screened
  contingencies.
MWn−1,AB − MWbase,A 1. Solve the power flow by removing the first line consid-
MWLODFAB = 100. (4)
MWbase,B ered in the n-1 contingency.
To generate the list of screened n-1-1 contingencies 2. Perform a time domain simulation and apply a 5-cycle
involving branch-branch outages, steps 1 to 7 excluding three-phase fault on the screened second line consid-
steps 3 and 6 as given in section III A are performed. ered in the n-1-1 case. The time domain simulation is
Steps 3 and 6 are excluded in this screening process as run only until the fault is cleared.
overloading of lines is a steady state issue and does not 3. Record the angular speed of generators in the area of
impact dynamic security assessment. Since system stability study at the end of the fault and compute the total
is of concern, change in the MW flow is a suitable screening kinetic energy gained by the system with respect to
parameter as it influences the rotor angles of the generators the center of inertia (COI) [25] of the area of inter-
in a system. A higher value of MWLODF indicates that only est. The total kinetic energy gained during the fault is

76 VOLUME 3, NO. 2, JUNE 2016

Mitra et al.: Systematic Approach to n-1-1 Analysis for Power System Security Assessment

given by opening of the faulted line weakens the transmission

    system. Such cases are likely to have stability issues
and warrant detailed time domain analysis.
X  X
ωcoi = Mi ω i   Mi  (5) II. Possibly Critical (PCR): A Contingency that results in
allgens allgens large kinetic energy and small 1Zthmax or small kinetic
1ωi = ωi − ωcoi (6) energy and large 1Zthmax is categorized as PCR.
X 1
KE = Ji (1ωi )2 (7) A large kinetic energy and small 1Zthmax implies that
2 although the fault is close to generation units, the post
allgens
disturbance network is strong. On the other hand, a
where large 1Zthmax and small kinetic energy indicates that
ωcoi is the angular velocity of the center of inertia although the post-fault network is weak, the fault is
ωi is the angular velocity of the ith generator electrically distant from large generation units. Contin-
Mi is the inertia constant of the ith generator gencies categorized as PCR may not be unstable under
KE is the kinetic energy gained by the system a given operating condition, but can become critical as
Ji is the moment of inertia of the ith generator the operating conditions change.
4. Remove the faulted line from service and compute the III. Non Critical (NCR): Contingencies that result in both
change in Zth as seen at the POIs of all generators/group small kinetic energy and small 1Zthmax are categorized
of generators in the region of interest. as NCR. No detailed time domain analysis is required
5. Record the maximum change in Zth and the corre- for these contingencies.
sponding bus number of the POI where it occurs. The Once the contingencies are categorized as CR, PCR and
maximum change in Zth will be referred to as 1Zthmax . NCR, they can be displayed on a kinetic energy vs. 1Zthmax
The stability of a power system during a fault depends on plot for ease of visualization. The visualization approach is
the kinetic energy gained by the system due to the fault and helpful as it enables a planner/operator to estimate the relative
the robustness of the post-disturbance network to absorb [24]. severity of the contingencies. Fig. 4 shows the three regions
This can be understood by examining the equal area criterion on the kinetic energy vs. 1Zthmax plot. It should be noted
for a one-machine infinite bus system. The idea has been
extended to multi-machine systems as well [25]. During a
fault, the ability of the network to export electrical power is
severely restricted causing the machine to accelerate. Once
the fault is cleared by opening the faulted line, the machine is
able to export electrical power and it decelerates. The stability
of the machine is dependent on its ability to decelerate in the
post-disturbance condition and reach a steady state.
The kinetic energy gained by the system during a fault
is computed by performing a short time domain simulation
with a three-phase fault applied on the concerned branch.
To estimate the ability of the system to decelerate, the change
in Zth seen into the system at the POI of the generator due to
opening of the faulted line is computed. The buses, which FIGURE 4. CR, PCR, and NCR, regions on the kinetic energy
are the POI of generating stations, are typically known to versus 1Zthmax .
planners beforehand. The change in Zth can be computed by
using sparsity oriented compensation methods [26]. A review that the boundaries of the three different regions are system
of the swing equation [24] indicates that a large change in dependent and hence, not defined explicitly. These bound-
Zth results in a substantial reduction in the peak of the post- aries can be defined ad hoc for a particular system and
fault swing curve as compared to the pre-faulted condition. subsequently tuned over time to increase the efficiency of the
A reduction in the peak of the post-fault swing curve limits classification process. The lack of an explicit definition of
the ability of the generator to decelerate, thereby making it the boundaries does not negate the classification process, as
prone to instability. Hence, these two factors are chosen to this process does not judge whether a contingency is unstable.
categorize contingencies based on their impact on stability. Rather, the classification process categorizes the contingen-
Based on the computed quantities the contingencies are cies based on their relative severity. A comprehensive stabil-
categorized into three distinct categories. Brief descriptions ity assessment is performed via TDS, which is a more reliable
of the three different categories are as follows method. The cases marked as CR and PCR should be assessed
I. Critical (CR): A Contingency is categorized as CR if by detailed TDS. Cases marked as NCR and LCR can be
both the kinetic energy and the 1Zthmax due to the assessed by early termination TDS [27], which is faster. The
contingency are high. A critical contingency implies contingencies should be assessed in the order CR and PCR
that the fault is electrically close to a generator and followed by NCR and LCR.

VOLUME 3, NO. 2, JUNE 2016 77

 IEEE Power and Energy Technology Systems Journal

C. n-1-1 CONTINGENCY ANALYSIS RESULTS FOR TABLE 7. Number of contingencies in different categories.
DYNAMIC SECURITY ASSESSMENT
MATLAB and GE PSLF [13] were used to perform the
n-1-1 contingency classification. As mentioned in
Section III-B, a 5-cycle three-phase fault was considered as
the disturbance on the system. An n-1 contingency analysis
was performed for DSA before proceeding to the four different categories. TDS were performed for the 204
n-1-1 analysis. 13 n-1 contingencies lead to islanding of screened n-1-1 contingencies. All the contingencies in the
generation units and all other contingencies were stable. The categories PCR and NCR were found to be stable. The 3 con-
13 contingencies were excluded as the first contingency for tingencies categorized as CR were found to cause instability.
the n-1-1 analysis. Table 8 presents the results of TDS for 8 selected contingen-
The screening process described in stage I of the proposed cies in categories CR and PCR. The 8 selected contingencies
method was used to screen candidate lines for simulating
the second outage. The MWLODF threshold for screening TABLE 8. Stability status of selected contingencies after
performing TDS.
candidate lines for second outage was set to 40%. Lowering
the MWLODF threshold did not result in screening of any
additional unstable contingency. Hence, the set threshold was
considered sufficient for the screening process. After the
second set of candidate lines are screened, a list of n-1-1
contingencies is created. The screened n-1-1 contingencies
are then classified appropriately by the classification process
described in stage II of the proposed method. Table 6 lists the
number of all possible n-1-1 contingencies and the number of
screened contingencies. The screening process substantially

(otg1-otg8) are marked in Fig. 5. The 8 contingencies were

TABLE 6. Metrics for the screening process in stage I.
chosen the cover the CR cases and the PCR cases, which
resulted in high kinetic energy or high 1Zthmax . The contin-
gencies in the region PCR are stable at this operating condi-
tion. This is because in these cases either the fault is far from
large generation or the post-disturbance network is strong
reduces the number of cases to be classified. As indicated enough to provide adequate deceleration to the accelerating
earlier the screening process of stage I is a conservative generators.
in nature. Not all screened contingencies are likely to have It is worthwhile to examine the contingencies categorized
dynamic issues. However, the screening process eliminates as PCR in further detail. As mentioned earlier, these con-
all the n-1-1 contingencies that are least critical from a tran- tingencies could become critical due change in operating
sient stability perspective. conditions. An example to illustrate such a case is as follows:
Fig. 5 shows the location of the screened n-1-1 contingen- Case: otg8 - outage of the lines 1518–1519 and 1211-1212
cies on the kinetic energy vs. 1Zthmax plot. The three different Outage of line 1518–1519 is considered as the n-1 outage.
A 5-cycle three-phase fault on line 1211-1212 and the ensu-
ing outage is considered as the n-1-1 event. Fig. 6 shows
a schematic one-line diagram of the area of interest. From
Fig. 5, it can be seen that the contingency otg8 results in

FIGURE 5. n-1-1 contingencies on a kinetic energy versus Zthmax

plot.

categories of contingencies are marked distinctly on the FIGURE 6. Schematic one-line diagram of region close to
plot. Table 7 lists the number of n-1-1 contingencies in the bus 1212.

78 VOLUME 3, NO. 2, JUNE 2016

Mitra et al.: Systematic Approach to n-1-1 Analysis for Power System Security Assessment

a large 1Zthmax , though the kinetic energy gained is com- V. PROCESSING TIME
paratively less. 1Zthmax for this contingency was recorded The processing time for the proposed n-1-1 analysis for both
at bus 1212, which is the POI of 12 generators CTG1 to SSA and DSA is presented in Table 10. The methods were
CTG 12, each rated at 44 MW. In the present scenario, only
4 generators are switched on producing 140 MW. Since the TABLE 10. Processing time of different subroutines for the
machines are smaller and have low inertia, the total kinetic proposed methods.
energy gained by the machines is not significantly high.
Opening the considered lines (marked crosses in Fig. 6) due to
the sequential outages, results in a large 1Zthmax . This is due
to the high impedance of the path 1212-1518-1160 (marked
with dotted line), through which the generation is connected
to the rest of the system in the post-fault condition. As
listed in Table 8, the contingency is stable for this operating
scenario.
The total generation capability at bus 1212 is 528 MW
(12 units, 44 MW each). As such, the generation level may implemented on an Intel Core i7-3770 @ 3.40 GHz processor
be increased under heavy load periods or due to the loss of with a 16 GB RAM.
generation at some other location in the system. Fig. 7 shows The following comments are in order.
the trace of the otg8 on the kinetic energy vs. 1Zthmax plot as 1. Step 1 needs to be performed first before proceeding
the generation at 1212 is increased. to the other steps. Step 1 is parallelizable if parallel
computing capability is available.
2. Step 2 is performed to generate the contingency list for
both SSA and DSA.
3. Step 3 and Step 4 are independent of each other and can
be done separately in any order.
4. Step 3 is parallelizable if parallel computing capability
is available.
5. Step 4a and Step 4b are independent of each other and
can be done separately in any order. Both step 4a and 4b
are parallelizable.
The timings presented in Table 10 are considering sequen-
tial processing due to the unavailability of parallel computing.
However, using the newer versions of PSLF [28], which have
FIGURE 7. Trace of otg8 on the kinetic energy versus Zthmax plot.
high performance computing (HPC) module [28] a substan-
tial reduction in the processing time is possible. It should
be noted that the methods discussed here are proposed for
Fig. 7 shows that as the generation increases, the con- n-1-1 analysis in the operations planning phase. Hence, the
tingency leaves the PCR region and moves towards the timings achieved in the presented report were sufficient for
CR region. This is because the added generators contribute to the intended purpose.
the kinetic energy. At a total output of 510 MW at bus 1212,
with 12 units online, otg8 results in both high kinetic energy VI. CONCLUSION
and 1Zthmax. At this operating condition, otg8 is in the CR This paper presents a systematic approach to n-1-1 con-
region and needs to be analyzed by detailed TDS. Table 9 list tingency analysis for a realistic large power system.
the results of the TDS performed for otg8 for five different An n-1-1 contingency screening and ranking method was
generation levels at 1212. proposed for SSA and an n-1-1 contingency screening and
From Table 9 it can be seen that otg8 leads to instability classification method was proposed for DSA. The main merit
when the generation at 1212 is increased. of these methods lies in the ease of implementation and the
TABLE 9. Results of TDS for changing generation levels at ability to screen critical contingencies reliably with reduced
bus 1212. computational effort. The screening methods limit the num-
ber of contingencies for which detailed analysis needs to be
performed. Ranking of the screened contingencies allows the
operator/planner to identify the most severe contingencies
and devise corresponding corrective measures from a steady
state perspective. Classification of contingencies from a sta-
bility viewpoint enables an operator to identify contingencies,

VOLUME 3, NO. 2, JUNE 2016 79

 IEEE Power and Energy Technology Systems Journal

which are unstable or may become unstable under changing [18] Y. Mansour et al., ‘‘BC Hydro’s on-line transient stability assess-
grid conditions. The accuracy of the proposed methods was ment (TSA) model development, analysis and post-processing,’’ IEEE
Trans. Power Syst., vol. 10, no. 1, pp. 241–253, Feb. 1995.
verified by a comparing the results with an exhaustive n-1-1 [19] V. Brandwajn, A. B. R. Kumar, A. Ipakchi, A. Bose, and S. D. Kuo, ‘‘Sever-
power flow analysis and n-1-1 TDS results. ity indices for contingency screening in dynamic security assessment,’’
The methods presented in this paper provide an approach IEEE Trans. Power Syst., vol. 12, no. 3, pp. 1136–1142, Aug. 1997.
[20] W. Li and A. Bose, ‘‘A coherency based rescheduling method for dynamic
to identify the potentially harmful contingencies quickly and security,’’ in Proc. 20th Int. Conf. Power Ind. Comput. Appl., May 1997,
reliably from the large number of possible contingencies. pp. 254–259.
In future work, the methods need to be implemented using [21] C. Fu and A. Bose, ‘‘Contingency ranking based on severity indices in
dynamic security analysis,’’ IEEE Trans. Power Syst., vol. 14, no. 3,
parallel computing infrastructure to identify the potential sav- pp. 980–985, Aug. 1999.
ings in processing time. Furthermore, the proposed methods [22] C. M. Davis and T. J. Overbye, ‘‘Multiple element contingency screening,’’
need to be tested on more actual power system cases, which IEEE Trans. Power Syst., vol. 26, no. 3, pp. 1294–1301, Aug. 2011.
[23] IEEE Standard for AC High-Voltage Circuit Breakers Rated on a Symmet-
will evaluate the efficiency of the methods. rical Current Basis—Preferred Ratings and Related Required Capabilities
for Voltages Above 1000 V, IEEE Standard C37.06-2009, Nov. 2009.
REFERENCES [24] A. A. Fouad and V. Vittal, Power System Transient Stability Analysis
Using the Transient Energy Function Method. Englewood Cliffs, NJ, USA:
[1] NERC. (Feb. 2008). System Performance Following Loss of a Single Prentice-Hall, 1992.
Bulk Electric System Element (Category B) (for This Particular Link), [25] D. Ernst, D. Ruiz-Vega, M. Pavella, P. Hirsch, and D. Sobajic, ‘‘A unified
accessed on Nov. 2015. [Online]. Available: http://www.nerc.com/files/ approach to transient stability contingency filtering, ranking and assess-
TPL-002-0.pdf ment,’’ IEEE Trans. Power Syst., vol. 16, no. 3, pp. 435–443, Aug. 2001.
[2] NERC. (Dec. 2007). Reliability Concepts, accessed on [26] O. Alsac, B. Stott, and W. F. Tinney, ‘‘Sparsity-oriented compensation
Nov. 2015. [Online]. Available: http://www.nerc.com/files/ methods for modified network solutions,’’ IEEE Trans. Power App. Syst.,
concepts_v1.0.2.pdf vol. PAS-102, no. 5, pp. 1050–1060, May 1983.
[3] L. L. Grigsby, Power Systems. Boca Raton, FL, USA: CRC Press, 2007. [27] G. C. Ejebe, C. Jing, J. G. Waight, G. Pieper, and F. Jamshidin, ‘‘Security
[4] A. A. Fouad, F. Aboytes, and V. F. Carvalho, ‘‘Dynamic security assess- monitor for online dynamic security assessment,’’ in Proc. 4th Int. Conf.
ment practices in North America,’’ IEEE Trans. Power Syst., vol. 3, no. 3, Power Syst. Control Manage., Apr. 1996, pp. 58–64.
pp. 1310–1321, Aug. 1988. [28] General Electric Int. Inc. (2013). GE Concorda PSLF Brochure
[5] K. Morison, L. Wang, and P. Kundur, ‘‘Power system security assessment,’’ HPC/SSTOOLS++, Schenectady, NY, USA, accessed on Feb. 2015.
IEEE Power Energy Mag., vol. 2, no. 5, pp. 30–39, Sep./Oct. 2004. [Online]. Available: http://goo.gl/hBK4uC
[6] G. C. Ejebe and B. F. Wollenberg, ‘‘Automatic contingency selection,’’
IEEE Trans. Power App. Syst., vol. PAS-98, no. 1, pp. 97–109, Jan. 1979.
[7] T. A. Mikolinnas and B. F. Wollenberg, ‘‘An advanced contingency selec-
tion algorithm,’’ IEEE Trans. Power App. Syst., vol. PAS-100, no. 2,
pp. 608–617, Feb. 1981.
[8] G. D. Irisarri and A. M. Sasson, ‘‘An automatic contingency selection
method for on-line security analysis,’’ IEEE Trans. Power App. Syst., PARAG MITRA (S’13) is currently working toward the Ph.D. degree in
vol. PAS-100, no. 4, pp. 1838–1844, Apr. 1981. electrical engineering at Arizona State University, Tempe, AZ, USA.
[9] N. M. Peterson, W. F. Tinney, and D. W. Bree, ‘‘Iterative linear AC power
flow solution for fast approximate outage studies,’’ IEEE Trans. Power
App. Syst., vol. PAS-91, no. 5, pp. 2048–2056, Sep. 1972.
[10] J. Zaborszky, K.-W. Whang, and K. Prasad, ‘‘Fast contingency evaluation
using concentric relaxation,’’ IEEE Trans. Power App. Syst., vol. PAS-99,
no. 1, pp. 28–36, Jan. 1980. VIJAY VITTAL (S’78–M’79–SM’87–F’97) received the Ph.D. degree from
[11] V. Brandwajn and M. G. Lauby, ‘‘Complete bounding method for AC con- Iowa State University, Ames, IA, USA, in 1982.
tingency screening,’’ IEEE Trans. Power Syst., vol. 4, no. 2, pp. 724–729, He is currently the Director of the Power Systems Engineering Research
May 1989. Center with Arizona State University, Tempe, AZ, USA.
[12] G. Irisarri, D. Levner, and A. M. Sasson, ‘‘Automatic contingency selection
for on-line security analysis—Real-time tests,’’ IEEE Trans. Power App.
Syst., vol. PAS-98, no. 5, pp. 1552–1559, Sep. 1979.
[13] GE PSLF User’s Manual, General Electric Int. Ltd., Schenectady, NY,
USA, Oct. 2012.
[14] M. A. El-Kady, C. K. Tang, V. F. Carvalho, A. A. Fouad, and BRIAN KEEL (M’07–SM’08) received the B.Sc. and M.Sc. degrees in
V. Vittal, ‘‘Dynamic security assessment utilizing the transient energy electrical engineering from the University of Illinois, Champaign, IL, USA,
function method,’’ IEEE Trans. Power Syst., vol. 1, no. 3, pp. 284–291, in 1988 and 1989, respectively.
Aug. 1986.
He is currently the Manager of Transmission System Planning with the
[15] H.-D. Chiang and J. S. Thorp, ‘‘The closest unstable equilibrium point
Salt River Project, Phoenix, AZ, USA.
method for power system dynamic security assessment,’’ IEEE Trans.
Circuits Syst., vol. 36, no. 9, pp. 1187–1200, Sep. 1989.
[16] G. C. Ejebe, G. D. Irisarri, W. F. Tinney, V. Vittal, and G. Cauley, ‘‘A sparse
formulation and implementation of the transient energy function method
for dynamic security analysis,’’ Int. J. Elect. Power Energy Syst., vol. 18,
no. 1, pp. 3–9, Jan. 1996. JENI MISTRY received the B.Sc. degree from Arizona State University,
[17] V. Chadalavada et al., ‘‘An on-line contingency filtering scheme for Tempe, AZ, USA.
dynamic security assessment,’’ IEEE Trans. Power Syst., vol. 12, no. 1, She is currently a Senior Engineer of Transmission Planning with the Salt
pp. 153–161, Feb. 1997. River Project, Phoenix, AZ, USA.

80 VOLUME 3, NO. 2, JUNE 2016