Reliability Evaluation in Transmission Systems
Reliability Evaluation in Transmission Systems
Reliability Evaluation in Transmission Systems
net/publication/306431265
CITATIONS READS
0 378
2 authors, including:
Joydeep Mitra
Michigan State University
168 PUBLICATIONS 1,868 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Joydeep Mitra on 12 December 2017.
1 Introduction
In reliability evaluation of the bulk power system (generation and transmission) two kinds of analyses are
used, depending on the extent and the purpose of the study: composite system analysis is used to perform
a detailed study of a system or a part thereof, while a multiarea study usually encompasses a much larger
system of interconnected utilities or control areas. Composite power system reliability evaluation is con-
cerned with the total problem of assessing the ability of the generation and transmission system to supply
adequate and suitable electrical energy to major system load centers. While in the multiarea reliability eval-
uation the transmission constraints are only indirectly considered, in the composite power system reliability
internal transmission limitations are directly modeled. These reliability studies can help in better represen-
tation of generation effects in transmission system reliability analysis and optimize relative investments in
generation and transmission systems, including dispersed generation.
It should be pointed out that multiarea and composite system reliability studies are both multinode and
similar in many ways. The major difference is in the transmission network modeling and because of the more
detailed network model, the composite system reliability model has many more nodes than the multiarea
model. In addition, the type of network flow models used in these types of studies is sometimes different.
Network flow model (transportation type) and DC flow methods are considered adequate for multiarea reli-
ability evaluation but DC flow or AC flow methods are considered more appropriate for composite system
reliability evaluation.
The methods used in composite system reliability evaluation could be either deterministic or probabilistic.
The main principle in the deterministic methods is to maintain adequate service under most likely outages
but to also accept some degradation of performance under low probability outages involving multiple gen-
eration and transmission facilities. The probabilistic methods model the factors affecting reliability more
comprehensively and may be analytical, Monte Carlo simulation, or hybrid.
We first discuss the component models and then the various methods of evaluation of reliability.
2 Component Models
For performing the reliability analysis of any system, the component models need to be properly defined.
The following components are considered in composite system reliability evaluation:
• Generating units
• Transmission lines
• Transformers
• Buses
• Circuit breakers
In addition, weather is an important component of this analysis and common mode failures of groups of
components need to be specified.
A two-state model is a special case of this model where there is no transition to state 2.
Up
1
Derated 2 Down
3
μ μ′
λ λ′
Down Down
1/S 1/N
Normal weather Adverse weather
Although weather can exist in several states it is typically assumed to exist either in the normal or the
adverse state. An important issue concerning weather is its extent of coverage area. When the system is
spread over a large area, at any given time, different regions may have different states of weather. An exact
treatment of this effect is difficult and some simplifications are needed. One approach divides the whole area
into regions. The weather in each region is characterized by mean duration of normal and adverse states.
The weather changes in different regions are assumed independent. Every line is assigned to a particular
region which really means that this line is predominantly affected by the weather in this region. It should be
noted that for parallel transmission lines just using failure rate, which is the average of normal and adverse
weather rates, can lead to error (Singh and Billinton, 1977).
1. Ground Fault. This refers to a fault in the circuit breaker itself. For this type of fault, the circuit breaker
is treated in the same manner as a transformer or a bus.
2. Failure to Open. The objective of a circuit breaker is to isolate the faulted component. Because of latent or
hidden faults in the breaker or the associated protection system, the breaker may not open when needed.
This may result in healthy components being isolated because of the operation of secondary zone pro-
tection. This failure mode is typically characterized by probability p in reliability modeling. This means
λ1
1U 1D
2U μ1 2U
μ2 μ2
λc
λ2 λ2
1U λ1
1D
2D μ1 2D
that when this breaker receives a command to open, there is a probability p that it may not respond. This
could be either due to a problem in the breaker or the associated protection system.
3. Undesired Tripping. It is also possible that a breaker may open without a command or fault. This can be
characterized as a rate and its effect will be an open line.
1. Contingency Ranking and Selection. This procedure consists of the following two steps (Mikolinnas and
Wollenberg, 1981; Endrenyi, 1978):
(a) Ranking of branch and/or generator outages for overload conditions and voltage problems using a
performance index (PI). This procedure will be briefly discussed later.
(b) Selection of contingencies from the ranked list using event probability or event depth cut off.
2. State Space Truncation. In this approach contingencies of only a certain order are considered. For
example, only up to second-order contingencies may be considered.
3. Implicit Enumeration Based on System Reliability Coherence. If a state is identified as a failure state, it
is assumed that all states resulting from this state by degradation of components are also failure states.
(Base-Case analysis)
(Select a contingency)
The use of min cut sets (Singh and Billinton, 1977; Liu and Singh, 2010) is actually an application of
this principle.
4. State Space Pruning Using State Space Decomposition. State space decomposition is a method of recur-
sively classifying the state space into coherent sets of functional and failure sets until all significant states
have been classified. The benefit of this method is that it deals with sets of states rather than individual
states. It can be made more efficient by using the method of pruning, which rapidly removes large parts
of the functional states before decomposition, thereby reducing computational time and effort (Singh and
Mitra, 1997).
5. Non-overlapping Failure Regions. Higher order outages are restricted to circuits in the same electrical
neighborhood, drastically decreasing the number of contingencies to be examined.
ĀF + G = D (1)
where
Ā = node-branch incidence matrix
F = circuit flow vector
G = generation vector
D = demand vector
G ≤ Gmax
F ≤ Ff
–F ≤ Fr
where
Gmax = vector of max available generation
Ff , Fr = vectors of forward and reverse branch capacities
Although this model is generally considered adequate for multiarea reliability analysis and has been used in
some commercial programs, for composite system reliability this model is not considered accurate enough.
B𝛉 + G = D (2)
G ≤ Gmax
F ≤ Ff
–F ≤ Fr
where
F = vector of flows fij
Gmax = vector of max available generation
Ff , Fr = vectors of forward and reverse branch capacities
The DC power flow model is generally felt to be a reasonable compromise between computational cost
and accuracy for planning studies. This is often used in regions where systems are strongly meshed and do
not have voltage problems.
In recent years, there has been renewed interest in development and deployment of power flow control
devices as part of the “smart grid” infrastructure. These devices include high-voltage DC lines (HVDC),
back-to-back AC–DC–AC ties, unified power flow controllers (UPFCs), as well as some other technologies
that allow increased control of power flow over specific lines independently of the impedances of these lines.
While detailed models of these controlled lines vary in complexity, linearized power flow models can
be adequately represented using the network flow model (1), because the flows on these lines are no longer
dependent on their impedances. Hence for these systems the linearized power flow models can be represented
as follows.
B𝛉 + Ācont Fcont + G = D (3)
where
Ācont = node-branch incidence matrix for the controllable lines only
Fcont = vector of real power flows on controllable lines.
Steps:
• Identification of failed operating constraints: line overloads, low/high bus voltages, generator reactive
power output violations, area net MW export violations
• Use of linear programming to determine optimal remedial actions
DC Mode:
• Phase shifter adjustment
• Generation re-dispatch
• Interruptible load curtailment
• Critical load shedding
AC Mode:
• MW and MVAR generation adjustment
• Gen bus voltage adjustment
• Phase shifter adjustment
• Transformer tap adjustment
• Switched cap/reactor
• Load curtailment
3.3 Contingency Ranking and Selection
Contingency ranking and selection is one of the techniques that has been used in a program developed by
Electric Power Research Institute. The goal of contingency selection techniques is to determine from the set
of all possible contingencies the subset that will cause system failure. No contingency selection method can
attain this goal perfectly; they can perhaps at best provide a subset that contains most contingencies causing
system failure.
One possible approach would be to rank contingencies by first solving each contingency using DC load
flow but it would be very time consuming. In a faster but less accurate method contingencies are ranked
approximately by severity based on a PI. The scalar function, called PI, is first defined to provide a measure
of system stress. Then a suitable technique is used for predicting ΔPI, that is, the change in PI when a
component is outaged. The ΔPI values for contingencies are then used to rank them in the order of severity.
Then AC or DC load flows are used to determine which of these ranked contingencies actually do cause
problems.
When a certain specified number of consecutive contingencies do not lead to system failure, the process is
stopped. The assumption here is that the remaining lower ranked contingencies will also not cause system
failure. This is not a foolproof method of ranking contingencies. It is possible that some severe contingencies
may be left out and also some not so severe contingencies may be ranked. Contingency ranking may be done
on the basis of either overload or voltage problems. A brief discussion is provided here for contingency
ranking based on overloads. A PI used for this purpose is described in EPRI report EL-2526 (1982):
( )n
∑ P𝓁
PI = W𝓁 (4)
𝓁 P𝓁
where
W𝓁 = weighting factor for circuit 𝓁
P𝓁 = real power flow on circuit 𝓁
P𝓁 = power rating of circuit 𝓁
n = an even integer, generally 2
Several approaches for finding ΔPI have been proposed. PI can generally provide a good measure of system
stress. In cases where load in one branch increases but that in others decreases, the PI may fail to recognize
overloads, resulting in a masking phenomenon. Masking can be reduced by increasing the exponent n but it
becomes difficult to solve ΔPI for n > 2.
This linear predictor has been found to produce reasonably good rankings for generating unit outages. An
implementation using DC power flow is shown below.
𝜕PI
= n𝜃̂i
𝜕Pi
where
𝜃̂i = ith component of vector ̂
𝛉
N = number of buses
̂
𝛉 = B−1 P ̂
B = N × N system susceptance matrix (as discussed in Section 3.1.2)
P̂ = vector of length N built by adding for every circuit 𝓁, an injection of P ̂𝓁 on
̂𝓁 on the from bus and −P
the to bus of circuit 𝓁 where
W𝓁 Bn𝓁 𝜃𝓁n−1
̂
P𝓁 = (6)
P̄ n𝓁
In other words, P̂ is the sum of 𝓁 vectors each of which contains P ̂𝓁 on the from bus and −P ̂𝓁 on the to
bus of circuit 𝓁, and 0 at all other buses. In Equation (6), 𝜃 𝓁 denotes the angle difference between the from
bus and to bus of circuit 𝓁. After the base case DC flow calculation, elements of B and the values of 𝜃 𝓁 are
available. One additional calculation is required to find ̂
𝛉.
where
X𝓁 = reactance of circuit 𝓁
where
X𝓁 = reactance of circuit 𝓁
P𝓁 = power flow on circuit 𝓁
i, j = from and to buses of circuit 𝓁
Poi , Poj = terms recognizing line charging and/or reactive sources and loads on buses i and j.
Let
F(x) = test applied to verify if state x is able to satisfy the load.
Now E(F) would be the true value of loss of load probability as it is computed by using the entire state
space X. In random sampling, x ∈ X are sampled from their joint distributions and then estimate of E(F) is
found using
1 ∑
NS
E(F) = F(xi ) (9)
NS i=1
where
NS = number of samples
xi = ith sampled value
F(xi ) = test result for ith sampled value
̂ V(F)
V(E(F)) = (10)
NS
Since V(F) is not known, its estimate can be used:
1 ∑
NS
̂
V(F) = ̂
F((xi − E(F)) 2
(11)
NS i=1
As explained later, the variance of the estimate is used to calculate the coefficient of variation that can be
used for testing the convergence of the simulation.
z = P(X ≤ xi ) (12)
1.0
P(X < x)
0
1 2 3
State xi
̂
V(E(F)) = variance of the estimator of E(F)
√
√
̂
V(E(F)) V(F)
COV = =√ (14)
̂
E(F) ̂
NSE(F)
From Equation 13 we can obtain the sample size needed to reach a certain level of COV:
V(F)
NS = (15)
̂
(COV ⋅ E(F)) 2
Some interesting conclusions can be drawn from Equation 15. It can be seen that sample size is not affected
by system size or complexity but by the accuracy required and the probability being estimated. The com-
putational effort depends on the sample size (NS) and the CPU time/sample. Thus, for a given COV and
the index although the number of samples is independent of the system size the CPU time is not, as the
time required for state evaluation for larger systems is higher than that for smaller systems. It should also
be remembered that most of the contribution to computation time comes from the state evaluation process.
The other factor that contributes to the sample size is the variance. If the variance is high then the sample
size is higher. This has led to using variance reduction techniques to reduce the computation time.
Then
E(Y) = E(F) (17)
then
V(Y) < V(F) (19)
If
E(F ′ ) = E(F ′′ ) = E(F)
then
E(F𝛼 ) = E(F)
To obtain negative correlation, use random numbers Zi to compute E(F′ ) and (1 – Zi ) to compute E(F′′ ).
z = P(Ui ≤ U)
= Fi (U)
A component with minimum time makes a transition and causes system transition. State evaluation is done
in the same manner as random sampling (Figure 6).
Fi (U) = P (U ≤ u)
z
u
Up time or time to failure
T = min {Ti }
If this T corresponds to Tp , that is, the pth component, then the next transition takes place by the change
of state of this component.
Step 2. The simulation time is now advanced:
tn+1 = tn + T
Tir = Ti − T
Step 6. From tn to tn+1 , the status of the equipment stays fixed and the following steps are performed:
a. The load for each node is updated to the current hour.
b. If no node has loss of load, the simulation proceeds to the next hour; otherwise the state evaluation
module is called.
c. If after remedial action all loads are satisfied, then the simulation proceeds to the next hour. Otherwise,
this is counted as loss of load hour for those nodes and the system. In addition, if in the previous hour
there was no loss of load, then this is counted as one event of loss of load.
d. Steps (a)–(c) are performed till tn+1 .
Step 7. The statistics are updated and the process moves to step 2.
Step 8. The simulation is continued till the convergence criterion is satisfied.
5 Cyber-Physical Considerations
In composite power system reliability evaluation, the cyber part including communication and protection
failures is generally assumed to be perfectly reliable. Thus the failure of a current-carrying component
such as a transmission line is assumed to result in the isolation of that component only. However, it has
been recognized that protection system hidden or undetected failures are common causes of multiple,
extended, or cascading outages (Singh and Patton, 1980a; Phadke and Thorp, 1996; Elizondo et al., 2001;
Yang et al., 2006; Yu and Singh, 2004). Studies such as in Yang et al. (2006), Yu and Singh (2004),
Singh and Patton (1980b), Billinton and Tatla (1983), Bozchalui, Sanaye-Pasand, and Fotuhi-Firuzabad
(2005), and Jiang and Singh (2011) indicate that protection system failure modes have significant effect
on reliability indices. Modern protection panels are equipped with multifunctional intelligent electronic
devices (IEDs) that are connected to communication networks. As indicated in Lei, Singh, and Sprintson
(2014) due to the variety of protection system architectures as well as the diversity of control and
communication mechanisms, it is hard to explicitly model protection systems with detailed configurations
simultaneously with the current-carrying components. Falahati, Fu, and Wu (2012) and Falahati and
Fu (2014) attempt to study the direct and indirect cyber-physical interdependencies, introduce some
mathematical terms and operations, and propose applications on small test systems including monitoring,
control, and protection features. These references provide valuable information on the impact of cyber
element failures on physical system reliability indices but do not provide a tractable method to deal with
the cyber-physical aspects of power systems. Lei, Singh, and Sprintson (2014) and Lei and Singh (2015)
propose a more systematic and scalable methodology for the overall analysis in a tractable manner with
the use of cyber-physical interface matrix (CPIM). The concept of CPIM provides a way of decoupling
the cyber analysis from the overall reliability evaluation. In fact, cyber analysis is done outside the
main simulation for composite system evaluation and then the results of analysis are interfaced through
CPIM. In Lei, Singh, and Sprintson (2014) and Lei and Singh (2015), a typical substation protection
system with detailed architecture is designed and analyzed as an example to illustrate the procedures of
obtaining a CPIM. The steps for using a CPIM in composite power system reliability evaluation are also
formulated.
7 Conclusions
This chapter provided a summary of the objectives and methods of composite system reliability evaluation.
These methods have been utilized for several applications in power system planning, such as (i) identi-
fication of transmission bottlenecks in the system under study, (ii) prioritization or optimization between
investments in generation and transmission, (iii) evaluation of the impact of transmission constraints in
economic and energy market analyses, and (iv) planning studies concerning integration of distributed
and variable resources. In modern grid operation practices where generation and transmission assets are
increasingly operated closer to their limits, the role of composite system analyses is becoming increasingly
important.
Related Articles
Evolution of Power Grids (Transmission);
Role of Energy Storage;
Cascading Failures in Power Systems;
Grid Codes in Power Systems with Significant Renewable-Based Generation;
Cybersecurity of SCADA within Substations;
Defining, Measuring, and Improving Resilience of Electric Power Systems;
Reliability Indices.
References
Bergen, A.R. and Vittal, V. (1999) Power System Analysis, 2nd edn, Prentice Hall, Upper Saddle River, NJ.
Billinton, R. and Tatla, J. (1983) Composite generation and transmission system adequacy evaluation including protection system
failure modes. IEEE Transactions on Power Apparatus and Systems, PAS-102 (6), 1823–1830.
Bozchalui, M.C., Sanaye-Pasand, M., and Fotuhi-Firuzabad, M. (2005) Composite System Reliability Evaluation Incorporating Protec-
tion System Failures. Proceedings of 2005 IEEE Canadian Conference on Electrical and Computer Engineering, May, pp. 486–489.
Elizondo, D.C., De La Ree, J., Phadke, A.G., and Horowitz, S. (2001) Hidden Failures in Protection Systems and their Impact on
Wide-Area Disturbances. Proceedings of IEEE Power Engineering Society Winter Meeting, January/February, vol. 2, pp. 710–714.
Endrenyi, J. (1978) Reliability Modeling in Electric Power Systems, John Wiley & Sons, Inc, New York, NY.
EPRI report EL-2526 (1982) Transmission System Reliability Methods, Volume 1: Mathematical Models, Computing Methods, and
Results.
EPRI report EL-5178 (1987) Composite System Reliability Evaluation Methods, June.
Falahati, B. and Fu, Y. (2014) Reliability assessment of smart grids considering indirect cyber-power interdependencies. IEEE Trans-
actions on Smart Grid, 5 (4), 1677–1685.
Falahati, B., Fu, Y., and Wu, L. (2012) Reliability assessment of smart grid considering direct cyber-power interdependencies. IEEE
Transactions on Smart Grid, 3 (3), 1515–1524.
Jiang, K. and Singh, C. (2011) New models and concepts for power system reliability evaluation including protection system failures.
IEEE Transactions on Power Systems, 26 (4), 1845–1855.
Kile, H. and Uhlen, K. (2012) Data reduction via clustering and averaging for contingency and reliability analysis. International Journal
of Electrical Power & Energy Systems, 43, 1435–1442.
Kim, H. and Singh, C. (2012) Comparison of Clustering Approaches for Reliability Simulation of a Wind Farm. International Confer-
ence on Power System Technology (POWERCON), Auckland.
Kim, H., Singh, C., and Sprintson, A. (2012) Simulation and estimation of reliability in a wind farm considering the wake effect. IEEE
Transactions on Sustainable Energy, 3, 274–282.
Lei, H. and Singh, C. (2015) Power system reliability evaluation considering cyber-malfunctions in substations. Electric Power Systems
Research, 129, 160–169.
Lei, H., Singh, C., and Sprintson, A. (2014) Reliability modeling and analysis of IEC 61850 based substation protection systems. IEEE
Transactions on Smart Grid, 5 (5), 2194–2202.
Liu, Y. and Singh, C. (2010) Reliability evaluation of composite power systems using Markov cut-set method. IEEE Transactions on
Power Systems, 25 (2), 777–785.
Mikolinnas, T.A. and Wollenberg, B.F. (1981) An advanced selection algorithm. IEEE Transactions on Power Apparatus & Systems,
100 (2), 608–617.
Mitra, J. and Vallem, M.R. (2012) Determination of storage required to meet reliability guarantees on island-capable microgrids with
intermittent sources. IEEE Transactions on Power Systems, 27 (4), 2360–2367.
Phadke, A.G. and Thorp, J.S. (1996) Expose hidden failures to prevent cascading outages. IEEE Computer Applications in Power,
9 (3), 20–23.
Singh, C. and Billinton, R. (1977) System Reliability Modelling and Evaluation, Hutchinson Educational, London.
Singh, C. and Gonzales, A.L. (1985) Reliability modeling of generation systems including unconventional energy sources. IEEE
Transactions on Power Apparatus and Systems, PAS-104 (5), 1049–1056.
Singh, C. and Kim, Y. (1988) An efficient technique for reliability analysis of power systems including time dependent sources. IEEE
Transactions on Power Systems, 3 (3), 1090–1096.
Singh, C. and Mitra, J. (1997) Composite system reliability evaluation using state space pruning. IEEE Transactions on Power Systems,
12 (1), 471–479.
Singh, C. and Patton, A.D. (1980a) Protection system reliability modeling: unreadiness probability and mean duration of undetected
faults. IEEE Transactions on Reliability, R-29 (4), 339–340.
Singh, C. and Patton, A.D. (1980b) Models and concepts for power system reliability evaluation including protection-system failures.
International Journal of Electrical Power and Energy Systems, 2 (4), 161–168.
Xu, Y. and Singh, C. (2014) Power system reliability impact of energy storage integration with intelligent operation strategy. IEEE
Transactions on Smart Grid, 5 (2), 1129–1137.
Yang, F., Meliopoulos, A.P.S., Cokkinides, G.J., and Dam, Q.B. (2006) Effects of Protection System Hidden Failures on Bulk Power
System Reliability. Proceedings of 2006 IEEE 38th North American Power Symposium, September, pp. 517–523.
Yu, X. and Singh, C. (2004) A practical approach for integrated power system vulnerability analysis with protection failures. IEEE
Transactions on Power Systems, 19 (4), 1811–1820.