You are on page 1of 5

A Risk-Based Reliability Method for N-1-1

Contingency Analysis
Muthanna Al-Sarray, Haider Mhiesan, Roy McCann Haitao Liao
Dept. of Electrical Engineering Dept. of Industrial Engineering
University of Arkansas University of Arkansas
Fayetteville, Arkansas, USA Fayetteville, Arkansas, USA

Abstract— There is a need for improved methods in conducting used to evaluate the operation of composite electric power
reliability studies of electric power systems. This is motivated by systems from an economic perspective in [8]. The results of
NERC TPL-001-4 for analyzing N-1-1 contingency events. In this optimal load flow analysis and load curtailment were employed
paper a new reliability analysis technique is developed. Referred when the system was unable to meet the load amount. Aleatory
to as a Risk-Based Reliability Method (RBRM), it is applied to a
68-bus, 16-machine IEEE benchmark to evaluate system
(failing components and repairing processes randomly) and
reliability under N-1-1 analysis. Factors include the transmission epistemic (limiting actual parameter values for probabilistic
line age, load flow under normal conditions and load flow under factors) were used in [9]-[11] to evaluate reliability with a
post-first-disturbance conditions. Results are also included where probabilistic Monte Carlo simulation. In [12]-[13] an approach
reliability factors are modified in order to improve the ability to utilizing steady-state load flow to analyze the reliability of
identify and rank the most severe and highly probable power systems included uncertainties in wind power was
contingencies. introduced. The probabilistic analysis in [14] illustrated that
increasing the number of surge arresters, and aging of cables,
Keywords— contingency analysis; protection systems;
and lines were factors influence reliability. [15] used weather
reliability; severity index.
forecast factors such as humidity, wind speed, temperature, and
I. INTRODUCTION their effects during maintenance periods as parameters that
impact power system equipment failure rates.
There is increasing awareness for considering multiple
consecutive equipment failures in reliability studies of electric In this research, a new method is proposed that calculates the
power systems. This is included in NERC TPL-001-4 standard reliability of a power system depending on a combination of
and it has recently motivated the development of new reliability variables such as transmission line ageing and power-flow. The
analysis methods for multiple contingency events. Traditionally, first factor is calculated based on the age of the particular
most regional transmission organizations (RTOs) and transmission line. The second factor is the transmission line
transmission operators have focused on N-1 contingency load-level obtained based on two conditions: the first is a load
analysis. This can be performed in a deterministic manner for flow under normal conditions when the system performs as an
real-time operational assessments by reliability coordinators. intact system; the second is the subsequent post-first fault
However, variable and distributed generation units are becoming condition when the system has a loss of one transmission line.
more pronounced in power systems and the need for N-1-1 These lead to developing a risk-based reliability method
analysis for them has led to consideration of probabilistic or risk- (RBRM). RBRM uses the ratio between the probability
based methods. The general goal for probabilistic methods is to weighted reliability levels at the first contingency event to
rank N-1-1 events based upon likelihood of occurrence and determine the final N-1-1 probability. This approach is
severity of the associated contingency sequence. In general, the developed for transmission networks and applied to transient
degree of power system reliability may be measured building on stability analysis of an IEEE 68-bus, 16-machine, 5-area
factors such as frequency, duration, and magnitude of adverse benchmark system [16].
impacts on the delivery of bulk electricity [1].
II. SEVERITY INDEX FOR N-1-1 CONTINGNECY EVENTS
There have been many probabilistic and risk-based methods
Fault conditions in power system components may occur due
developed to analyze reliability of electric power systems. A
to spontaneous failures (such as a line-to-ground fault) or
risk-based approach can help planners and transmission
misoperation events such as an anomalous relay tripping of a
operators to mitigate potential fault conditions via simulating
circuit breaker. In planning studies, power system engineers
different scenarios in advance. As examples, [2]-[7] account for
analyze for possible fault occurrences in order to evaluate the
factors that affect the reliability and security of power systems.
ability of automatic protective functions to minimize
Those factors include protection systems having hidden failures
disturbances and restore power. Contingency analysis (CA) of
present in relays, transducers and circuit breaker trip
abnormal cases in electric systems has been used in power
mechanisms. Probability of failure and frequency of failure were
system planning. CA has accounted for the impact of instrument
outages [17]-[18]. In this work, N-1-1 contingency events are
This research was funded in part by the National Science Foundation
Industry/University Cooperative Research Center on Grid-connected Advanced analyzed through simulation of an IEEE 68-bus transient
Power Electronics Systems (GRAPES) under grant number 0934390.

978-1-5090-3270-9/16/$31.00 ©2016 IEEE

Authorized licensed use limited to: Universiti Malaysia Perlis. Downloaded on June 03,2021 at 13:50:47 UTC from IEEE Xplore. Restrictions apply.
stability benchmark power system [16]. A severity index for
assessing the impact of N-1-1 fault condition is used similar to
that detailed in [19]. The severity index is calculated for a power
system with n-generators by forming a vector of rotor angles =

[ߜଵ ߜଶ ߜଷ … ߜ௡ ሿ where ߜ௜ (t) is the time-function defined for
௧௛
the ݅ generator rotor angle for the transient response to an N-
1-1 contingency sequence. The maximum absolute value of a
time function over a bounded time is found from evaluating
ԡߜ௜ ԡஶ ൌ ‫݌ݑݏ‬ȁߜ௜ ሺ‫ݐ‬ሻȁǣ ‫ א ݐ‬ൣͲǡ ‫ݐ‬௙ ൧ where ‫ݐ‬௙ is the final time for
evaluating the transient response [19]. The resulting maximum
(worst-case) rotor angle deflection ԡߜ௜ ԡஶ is defined as the
severity index for that particular N-1-1 sequence.
III. RISKED BASED RELIABILITY METHOD (RBRM)
The approach in this paper is developed to evaluate the
reliability of power systems depending on three factors: age of
transmission line; load flow under normal conditions; and load
flow under post-first disturbance conditions. A probabilistic
method is used to select a transmission line randomly and
calculate the load flow for its normal and post-fault cases. The
proposed approach is shown in Fig. 1.
The age of the transmission lines is assumed as a proxy for
probability of failure in order to obtain a reliability measure for
each case. Because of the lack of historic data for the
benchmark test system, a hypothesized age estimation is done Fig. 1. Reliability flow chart.
in order to complete the reliability calculations. The values for
line age are given in seconds in order to relate standard These values can be adjusted to evaluate the behavior of the
electrical SI units, but in the actual practice the line’s age ܶ௔ is power system to possible design changes, such as replacing or
given in thousands of hours. For analysis purposes in this upgrading transmission lines.
research, the value of ܶ௔ has a normalizing factor applied to it
in order to provide a measure of the age of each line relative to Using ߟଵ௜ andܶ௜ , the reliability of ith line under normal
one another. This avoids unreasonably large numerical values statues can be given as in [20]-[23]
(in seconds) given that the assumed line ages are all greater than ഁ
ܴଵ௜ ൌ ݁ ିሺ்೔Ȁఎభ౟ሻ , (3)
10 years. For obtaining the line’s relative age ܶ௔ ǡa
normalization factor ܶ௥ is selected arbitrarily to be equivalent where ܴଵ௜ is the reliability of line i at the normal condition and
to 10 years and the scaled value ܶ௜ for the line age ܶ௔ is ܶ௜ is the age of line i, while ߚ is a shape parameter.
ܶ௜ ൌ  ܶ௔ Ȁܶ௥ . (1) After the first reliability calculations, the same steps would be
The resulting values are summarized in Table. I. utilized to calculate the same variables but under different
After the line ages’ normalization, the current that passes conditions. Via the load flow after the first event, ߙ଴ and ߙଵ ,
through the line for the non-fault case is calculated and used new calculations would be shown in [20]-[23]
along with the line age to determine the probability of a failure
occurring on that line. These factors are used for the ith line as ߟଶ௜ ൌ ߙ଴ ݁ ିఈభ‫כ‬ௌమ௜ , (4)
[20]-[23] where ߟଶ௜ is a characteristic life parameter under the post-first
fault case and ܵଶ௜ is the load flow in the post-first fault condition
ߟଵ௜ ൌ ߙ଴ ݁ ିఈభ‫כ‬ௌభ೔ , (2)
at line i.
where ߟଵ௜ is a characteristic life parameter (mean life) under the
normal case, ߙ଴ and ߙଵ are the coefficients in the life-stress TABLE I. THE AGE OF SOME LINES IN SECONDS
relationship parameters (the scale parameters of the distribution
function an) ߙ଴ and ߙଵ can accelerate the life testing of lines to Line Number Bus From – Bus To ࢀ࢏
analyze its reliability under different conditions.ܵଵ௜ is the load
flow in the normal condition at line i. ߙ଴ and ߙଵ are calculated 1 27 - 53 190
based on the ratio between values of load flows for lines i and j 22 48 - 47 901
in normal condition after comparing them with the maximum 45 45 - 39 552
value in that case. Calculations of ߙ଴ and ߙଵ depended on 54 23 - 22 206
assumptions that all lines have the same conductor, and a line 63 17 - 36 536
with low load flow would have high reliability and vice versa.

Authorized licensed use limited to: Universiti Malaysia Perlis. Downloaded on June 03,2021 at 13:50:47 UTC from IEEE Xplore. Restrictions apply.

߬௜ ൌ  ߟଶ௜ ඥሺെ Ž‘‰ሺܴଵ௜ ሻሻ, (5)
where ߬௜ is the time that line i can carry load flow ܵଶ௜ normally.
Equation (4) shows the approach to calculation of ߬௜ when the
line carries the different current after disturbances with taking
in account that ߙ଴ and ߙଵ are constant.߬௜ would be higher than
ܶ௜ because ߬௜ comes from ܶ௜ plus the time of the post-first fault’s
current. After getting ߬௜ , the reliability of line after the first fault
can be obtained by using ߬௜ and ܶ௙ and given in [20]-[23]

ܴଶ௜ ൌ ݁ ିሺఛ೔ା்೑Ȁఎమ೔ ሻ , (6)
where ܴଶ௜ is the reliability of line i after the first fault and ܶ௙ is
the time of first fault. From previous steps, the reliability of
each line could be modified in the formula
ܴଶଵ௜ ൌ ܴଶ௜ Ȁܴଵ௜ . (7)
where ܴଶଵ௜ is the final reliability of line i given by the first fault
on its system. By using this analysis, the prediction of an N-1-
1 contingency event can be obtained by using an inverse
relationship between reliability and probabilityሺܴ ൌ ͳ െ ௥ ሻ. Fig. 3. Reliability of 68-bus system with ᅈ=1.

IV. SIMULATION RESULTS RBRM to predict the most severe cases.


In this context, the severe cases where generators lose
Through using the 68-bus IEEE benchmark model in [16],
synchronism have small reliability values due to the
the severity index (SE) is calculated for showing the severe
exponential function that is utilized in the reliability
cases of double contingencies [19]. Fig. 2 shows the SE for all
assessment. RBRM is applied to evaluate the reliability and
possible N-1-1 contingencies for line outages in the 68-bus
predict which line will be most likely to have the second fault.
system. These cases are sorted depending on (x, y, z)
Selection of the values of ߙ଴ and ߙଵ is conducted after choosing
coordinates, where the x-axis is the number of line which has
a specific line and using its load flow with a specific value of ߚ
the first fault, the y-axis is the number of line which has the
via (2) to (5). This process is accomplished through adopting
second fault, and z-axis is the value of SE. Each case in Fig. 2
two values of load flow (both normal and post-first fault
has a different value of SE that depends upon the load flow and
conditions) for a specific line. This line is selected according to
maximum rotor angle deflection. Cases that have high SE are
its load flow value. Both values of load flow have been
visually emphasized in Fig. 2 to draw attention to those N-1-1
employed for different moments (ܶ௜ and ߬௜ ). By applying (2) to
contingencies that justify increased attention fo corrective
(5), the values of ߙ଴ and ߙଵ are obtained. Subsequently, the
action. Depending on these cases and their computed reliability,
values of ߙ଴ , ߙଵ becomes constants for each ߚǤ To analyze the
system planners and operators can mitigate the N-1-1 effects by
system equipment upgrades or design changes [19]. The RBRM of the system,ߙ଴ , ߙଵ and ߚ are varied (ߚ is assumed that
purpose of the SE’s calculation is to use it with the results of varied from 1 to 10). This range provides the ability to observe
the cases that have the worst impacts on the overall system
reliability. These changes are shown in Fig. 3 through Fig.5. In
these figures, the x-axis represents the first fault case, the y-axis

Fig. 2. Severity index (SE) for N-1-1 cases. Fig. 4. Reliability of 68-bus system with ᅈ=3.

Authorized licensed use limited to: Universiti Malaysia Perlis. Downloaded on June 03,2021 at 13:50:47 UTC from IEEE Xplore. Restrictions apply.
power flow would be increased in some of the remaining lines,
thereby generating cases of lowered reliability for those lines
most impacted by increased power flow (i.e., reduced thermal
margins).
On the other hand, having contingency cases with small
values of probability distribution function of RBRM, due to the
value of the shape parameter ȕ, makes recognizing the lines that
have low reliability (high failure probability) more complex.

When the term െሺ ೔ ሻ is increased by having high ȕ in (3) and
ఎభ೔
(6), the value of the distribution function would be decreased.
After changing the value of ȕ, shown in Figs. 4 and 5 with ȕ
equal 3 and 7, respectively, the probability distribution function
of RBRM is changed. Selection the optimal value of ߚ could
play a favorable role in distinguishing the cases with high
reliabilities. In Fig. 4 where ߚ equal to 1, the reliability of
different cases almost have the same value and that leads to
Fig. 5. Reliability of 68-bus system with ᅈ=7. difficulty in finding the line that would be likely to have a next
fault. In Fig. 5, where ߚ is equal to 3, the different values of
represents the transmission line number (identifier label), the z- reliability between cases can be recognized easily and this
axis is the value of the reliability of the (‫ ݔ‬௧௛ ǡ ‫ ݕ‬௧௛ ) associated becomes more clear when ߚ is equal to 7 as displayed in Fig. 5.
transmission line and fault case. However, if ߚ is further increased, the distinction between cases
As illustrated in Fig.3 through Fig. 5, the reliability for some would be difficult to discern. To indicate the challenge of
cases is affected by factors such as the amount of load flow and distinguishing the change in reliability for cases with high ߚ,
ȕ. Increased power rating leads to distinguishing between (8) is developed to identify the values of reliability in five cases
normal cases having high values of reliability and abnormal (ߚ has a range between 1 and 10) through the reliability values,
cases having reduced reliability due to the differences between by using the definition of collection as
the values of currents for normal and abnormal conditions. The
high value of active power flows in this system under normal ܵ‫݉ݑ‬ሺߚሻ ൌ σ௞௠ୀଵ ܴଶଵ೘೙ Ǥ (8)
conditions would not exceed 50 MW, while the active power ௡ୀଵ
passes under abnormal conditions may exceed hundreds of where m and n are the index numbers of transmission lines.
MW. This is interpreted as the line i having a high load flow These values are tabulated in Table II where the cases that have
(apparent power) passing through it; the reliability of it would high collection imply that the determination of the line’s
be low compared with lines that have low power flow, due to reliability could be computed in a more effective manner. Cases
the natural exponential function producing the reliability that have low collection mean it is difficult to distinguish
evaluation and depending on the amount of power flow in between the variation of the line’s reliability. Building on the
addition to values of ߙ଴ and ߙଵ . The impacts of the value of ߙ଴ results shown in Fig. 3 to 5 and table II, ȕ equal to 3 or 5 would
and ߙଵ come from their roles in calculating the characteristic be optimal values for prediction the next contingency event
life parameter (ࣁ) as indicated in (2) and (4). Also, when a according to RBRM.
generator loses synchronism due to severe disturbances, the In the previous process, each ȕ has specific values of ߙ଴
and ߙଵ and that does not show the impact of varying one factor
on others. Therefore, there is a need to determine how the
probability of a disturbance affects reliability if ȕ has different
value of ߙ଴ and ߙଵ . Revision of the values of ȕ for a range
between 1 to 7 while keeping ߙ଴ and ߙଵ constant was evaluated
in four different scenarios (ȕ is equal to 1, 3, 5, and 7 for each
respective scenario) and shown in Fig. 6. In this figure, the sum
SE of all potential N-1-1 cases is obtained for all scenarios after
changing ȕ between 1 and 7. After that, ߙ଴ and ߙଵ are varied
and then ȕ is altered again 1, 3, 5 and 7 (for the four scenarios).
Fig. 6 shows how the value of ȕ impacts the distribution
function of RBRM.
It was observed in the previous N-1-1 scenarios that
although ߙ଴ and ߙଵ are constants, distinguishing between cases
with high failure probability are enhanced depending on the
value ofߙ଴ . To consider this further, Fig. 6 displays the impact
of keeping ߙ଴ constant and varying ȕ. An increased value of ߙ଴
Fig. 6. Collection with respect to ᅈ under different scenarios. keeps the SE almost the same for all cases when ȕ is changed

Authorized licensed use limited to: Universiti Malaysia Perlis. Downloaded on June 03,2021 at 13:50:47 UTC from IEEE Xplore. Restrictions apply.
from 1 to 7 in the four scenarios, as evident specifically in [6] F. Yang, A. P.Sakis Meliopoulos, G. J. Cokkinides, Q. B. Dam, “Effects
of Protection System Hidden Failures on Bulk Power System Reliability,”
scenarios 1 and 2. In scenario 1 and 2, ߙ଴ has relatively high 38th North American Power Symposium( NAPS 2006), Carbondale, IL,
values, while scenario 3 and 4 show that impact of ȕ variation 17-19 Sept. 2006, pp.517-523.
would be large if ߙ଴ value is set to a low value. [7] F. Yang, A. P.Sakis Meliopoulos, G. J. Cokkinides, Q. B. Dam, “Bulk
Power System Reliability Assessment Considering Protection System
According to this analysis, one concludes that choosing ߙ଴
Hidden Failures,” Bulk Power System Dynamics and Control - VII,
and ȕ values plays a significant role in the results of relaibility Revitalizing Operational Reliability Symposium(IREP2007), Charleston,
analyss using RBRM. Also, the selections of ߙ଴ and ȕ help to SC, August 19-24, 2007 , pp.1-8.
identify the events which have increased failure probability. [8] C. T. Su, J. J. Wong, C. J. Fan, “System and Load Points Reliability
Moreover, the values of power flow, ȕ, ߙ଴ ǡ and ߙଵ impact the Evaluation for Electric Power Systems,” 1st Annual IEEE Systems
Conference, Waikiki Beach, Honolulu, Hawaii, April 9-12, 2007, pp.1-7.
reliability evaluation such that increasing these factors enable
[9] R.N. Allan, R. Billinton, A. Gawad, “The IEEE Reliability Test System -
to better identification of post-first fault cases that have more
Extensions to and Evaluation of the Generating System”, IEEE
severe outcomes. Transactions on Power Systems, vol. PWRS-1, no. 4, pp. 1-7, Nov. 1986.
[10] R. Billinton, W. Wangdee, “Delivery Point Reliability Indices of a Bulk
TABLE II. THE DIFFERNCES OF RISKED-BASED RELIABILIT METHOD Electric System Using Sequential Monte Carlo Simulation”, IEEE
(RBRM) Transactions on Power Delivery, vol. 21, no. 1, pp.345-352, Jan. 2006.
[11] R. Billinton, D. Huang, “Aleatory and Epistemic Uncertainty
ᅈ ࢻ૙ ࢻ૚ Sum(ȕ) Considerations in Power System Reliability Evaluation,” Proceedings of
the 10th International Conference on Probabilistic Methods Applied to
1 16786.79 0.1512 3659.24 Power Systems( PMAPS '08), Rincón, Puerto Rico, 25-29 May 2008,
pp.1-8.
3 2637 0.0599 2800.27 [12] R. Billion, W. Wangdee, “Reliability-Based Transmission Reinforcement
Planning Associated With Large-Scale Wind Farms,” IEEE
5 1977 0.0636 1862.47 Transsactions on Power systems, vol. 22, no. 1, pp.34-41, Feb. 2007.
[13] X. Han, M. Mu, W. Qin, “Reliability Assessment of Power System
7 1700 0.0576 1156.44 Containing Wind Farm Based on Steady-state Power Flow,” IEEE 11th
International Conference on Probabilistic Methods Applied to Power
Systems (PMAPS), Singapore, Singapore, 14-17 June 2010, pp.756-760.
10 1513.79 0.0527 650.9
[14] Y. Damchi, J. Sadeh, “Effect of Combined Transmission Line (Overhead
Line/Cable) on Power System Reliability Indices,” 4th International
V. CONCLUSION Power Engineering and Optimization Conf. (PEOCO2010), Shah Alam,
Selangor, MALAYSIA: 23-24 June 2010, pp.59-63.
Studying the reliability of power systems is essential in
[15] X. Du, W. Liu, “Evaluation of power system reliability based on the
helping reliability coordinators and system planners to decide maintenance state,” 4th International Conference on Electric Utility
upon remedial actions in critical situations. The proposed Risk- Deregulation and Restructuring and Power Technologies (DRPT),
Based Reliability Method (RBRM) is utilized to evaluate the Weihai, Shandong, China, 6-9 July 2011, pp. 1016-1020.
system reliability after a first fault and to predict the probability [16] B. C. Pal, and A. K. Singh, “IEEE PES Task Force on Benchmark
of an N-1-1 event. It was shown how RBRM enhances the Systems for Stability Controls Report on the 68-Bus, 16- Machine, 5-Area
System,” Version 2.0, July, 2013.
ability to distinguish between the severities of various N-1-1
[17] V. J. Mishra, M. D. Khardenvis, “Contingency Analysis of Power
contingencies. The results of the research indicate that RBRM System,” 2012 IEEE Students’ Conference on Electrical, Electronics and
would improve the ability of reliability coordinators to identify Computer Science, Bhopal, India, 1-2 March 2012, pp. 1– 4.
weak elements in a power system and thereby provide a guide [18] M. Chen, “Dynamic Contingency Re-definition in Power System Security
to transmission planners in prioritizing system upgrades. Analysis,” 4th International Conference on Electric Utility Deregulation
and Restructuring and Power Technologies (DRPT), Weihai, Shandong,
REFERENCES China, 6-9 July 2011, pp. 63-66.
[1] Measurment Practices for Reliability and Power Quality. A toolkit of [19] M. Alsarray, M. Saadeh, R. McCann, “A Risk-Based Planning Method
reliability measurments practices. Retrieved from for N-1-1 Contingency Analysis of Transmission Systems,” North
http://www.science.smith.edu/~jcardell/Courses/EGR325/Readings. American Power Symposium ( NAPS 2015), Charlotte, NC, Oct 4-6 2015,
pp. 1-6.
[2] Z. Bao-hui, W. Li-yong, Z. Wen-hao, Z. De-cai, “Implementation of
Power System Security and Seliability Considerieng Risk under [20] M. Bourguignon, R. B. Silva, G. M. Cordeiro, “The Weibull-G Family of
Environment of Electriciy Market,” IEEE/PES Transmission and Probability Distributions,” Journal of Data Science, vol. 12, pp. 53-68,
Distribution Conference and Exhibition, Asia and Pacific, Dalian, China, 2014.
2005, pp.1-6. [21] P. Jiang, Y. Xing, X. Jia, B. Guo, “Weibull Failure Probability Estimation
[3] M. P. Ehavaraju, R. Eillinton, N. D. Reppen,“Requirements for Based on Zero-Failure Data,” Mathematical Problems in Engineering,
Composite System Reliablity Evaluation Models”, IEEE Transaction on vol. 2015, no. 681232, pp. 1-8, 2015.
Power Systems, vol. 3, no.1, pp. 149 – 157, 1988. [22] G. Huang, W. Lin, Q. Niu, “Risk Analysis Model of Automobile Defect
[4] Y. Sun, L. Cheng, H. Liu, S. He, “Power System Operational reliability Based on Weibull,” International Journal of Hybrid Information
evaluation Based on Real-time Operating State,” 7th International Power Technology, vol.9, no. 1, pp. 353-366, 2016.
Engineering Conference(IPEC2005), Singapore, Singapore, Vol. 2, Nov. [23] M. S. Khan, R. King, “Transmuted Modified Weibull Distribution: A
29-Dec. 2 2005, pp.722-727. Generalization of the Modified Weibull Probability Distribution,”
[5] A. G. Phadke, J. S. Thorp, “Expose Hidden Failures to Prevent Cascading European Journal of Pure and Applied Mathematics, vol. 6, no. 1, pp. 66-
Outages,” IEEE Computer Application in Power, Vol. 9, No. 3, pp.20-23, 88, 2013.
1996

Authorized licensed use limited to: Universiti Malaysia Perlis. Downloaded on June 03,2021 at 13:50:47 UTC from IEEE Xplore. Restrictions apply.

You might also like