Module PE - PAS.U22.5: Composite System Reliability
Module PE - PAS.U22.5: Composite System Reliability
Module PE - PAS.U22.5: Composite System Reliability
Module PE.PAS.U22.5
Composite system reliability
U22.1 Introduction
We briefly introduced composite system reliability in module U17,
(also known as HL-II or G&T adequacy analysis), where we
identified two attributes that characterize most of the approaches
developed to date. These attributes are:
Method of representing stochastic nature of the operating
conditions: By “operating conditions,” we are referring to the
basecase network configuration (topology and unit
commitment) together with the loading and dispatch. The
methods include nonsequential and sequential enumeration.
Method of representing stochastic nature of contingencies: The
methods include contingency enumeration and Monte Carlo.
These methods are summarized in Table U22.1.
Table U22.1: HL-II Evaluation approaches
No
SYSTEM PROBLEMS?
Yes
Use linear programming
to obtain minimum gen
SIMULATE REMEDIAL ACTIONS redispatch cost, reactive
support cost, or load
Yes curtailment.
No
SYSTEM PROBLEMS?
No
CONTINGENCIES DONE?
Yes
No
OPERATING CONDITIONS DONE?
ACCUMULATE:
RELIABILITY INDICES x LOAD PROBABILITIES
For example, below is the Y-bus and the B’ matrix for a network
having no resistance.
30 10 10 10
10 20 10 0
Y j
10 10 30 10
10 0 10 20
20 10 0
B' 10 30 10
0 10 20
Although the DC power flow equation P=B’θ provides the ability
to compute the angles, it does not directly provide the line flows. A
systematic method of computing the line flows is:
P B ( D A)
where:
PB is the vector of branch flows. It has dimension of M x 1,
where M is the number of branches in the network. Branches
are ordered arbitrarily, but whatever order is chosen must also
be used in D and A.
Module PE.PAS.U22.5 Composite system reliability 5
1 2
2
1
5 3
4
4 3
1 0 0 - 1
1 -1 0 0
A 0 1 1 0 (5 rows because 5 branches; 4 columns because 4 buses.)
0 0 1 1
1 0 1 0
for all N-1 and N-2 contingencies. The problem with this
approach is that it may spend a great deal of computation on
contingencies that have little effect on the resulting indices, or,
on the other hand, it may miss some higher-order contingencies
that do significantly contribute to the reliability indices.
Limit the probability: Here, one analyzes all contingencies for
which their probability exceeds a certain threshold. This is
slightly more refined than limiting the order, and it is subject to
the same problems.
Based on severity: The principle here is to avoid analyzing
contingencies that do not result in a problem. Some type of
screening approach is normally involved, using either the DC
load flow or the linear sensitivities. It is convenient to also
associate with each contingency k screened a performance index
PIk given by:
n
Pj
Nb
PI k
j 1
Pj ,max
(U22.3)
on the assumption that the operator will make the least-cost control
adjustments in order to alleviate any overloads; such an
assumption can be accommodated in the formulation to follow.)
The most comprehensive algorithm available for implementing
remedial actions is the optimal power flow (OPF). The advantage
of the OPF is that it provides the ability to simulate operator action
for overload, low voltage, and voltage instability. However, it is
computationally expensive, although computing power today via
combination of high-end and parallel processing makes it quite
practical, even for large systems. We provide a simpler approach
which only allows remedial actions for overloads. One may
consider it to be an approximation to the OPF.
Define Pgi and Pgi to be an increase or decrease, respectively,
in the generation at bus i, and Pdj to be the load curtailment at
bus j, where all three variables are always nonnegative.
Also define Ki and Kj to be weights associated with the bus i
generation change and the bus j load curtailment, respectively.
The bus i generation change weight Ki should be chosen based
on dispatchability (infinity if not dispatchable) and ramp rates
(the higher the ramp rate, the smaller the Ki) for each plant,
indicating that the faster the plant can move, the more we will
rely on it in order to implement the desired remedial action as
soon as possible.
The bus j load curtailment weight Kj should be chosen based on
whatever contracts are in place with the loads. It is typical that
Kj>>Ki indicating that load curtailment is typically much less
desirable than generation redispatch.
Module PE.PAS.U22.5 Composite system reliability 12
subject to:
ng ng nd
P P P
i 1
gi
i 1
gi
j 1
dj 0 (U22.4b)
ng ng nd
P Ck ,i P Ck ,i P Ck ,i Pdj Pkmax k (U22.4c)
k
0
gi
gi
i 1 i 1 j 1
1 1 1 1 p1p2p3
2 1 1 0 p1p2q3
3 1 0 1 p1q2p3
4 1 0 0 p1q2p3
5 0 1 1 q1p2p3
6 0 1 0 q1p2q3
7 0 0 1 q1q2p3
8 0 0 0 q1q2q3
The binary tree for this system represents the different states of
Table U22.2, and is illustrated in Fig. U22.3.
Module PE.PAS.U22.5 Composite system reliability 15
p1 q1
Component 1
p2 q2 p2 q2
Component 2
p3 q3 p3 q3 p3 q3 p3 q3
Component 3
p1 p2 p3 p1 p2 q3 p1 q2 p3 p1 q2 q3 q1 p2 p3 q1 p2 q3 q 1 q2 p3 q1 q2 q3
LOLP Pr[s]
sFailedStates
(U22.5)
The problem is, of course, that the tree can become very large. For
example, if we have just 100 components, which would of course
be a very small system, we would have 2100=1.271020 terminal
nodes (or states).
U22.5.2 Obtaining a lower bound
We turn now to a general case corresponding to a system with N
components, but we assume that each one is modeled with only 2
states.
We define Prj[F] as the additional failure probability due to
knowledge that contingency j fails. Note that contingency j does
not simply indicate that “component j fails,” i.e., contingency j
simply indicates the jth contingency, which may or may not be
comprised of multiple component failures.
Module PE.PAS.U22.5 Composite system reliability 16
pk qk
Component k
p1 q1 p1 q1
p2 q2 p2 q2
Fig. U22.4: Binary tree for illustrating system states with circuit k
at the tree-top
Now let’s assume that all terminal nodes stemming from qk are
failed states. Under this assumption, we have that:
Module PE.PAS.U22.5 Composite system reliability 17
Prj [ F ] Pr[s]
s[ qk stem]
(U22.7)
LOLP
j Failed States
Prj [ F ]
APP (U22.9)
N
Pr j [ S ] qk pi Pr j [ S ]
APP
i 1, (U22.10)
ik
LOLP
j SuccessStates
Prj [ S ]
APP (U22.12)
N
Pr j [ F ] qk pi
APP
i 1,
ik
Pr[ F ] Pr[ F ] Pr j [ F ]
APP
b. ElseIf j=success, then
i. Compute success probability approximation
N
Pr j [ S ] qk pi
APP
i 1,
ik
Pr[ S ] Pr[ S ] Pr j [ S ]
APP
Then we have that the actual LOLP is bounded as illustrated in
Fig. U22.5.
1.0
Pr[F] 1-Pr[S]
LOLP resides
within this interval
Fig. U22.5: Illustration of bounds on LOLP
Module PE.PAS.U22.5 Composite system reliability 20
Ng Mj
s Pkj s P
T
min kj
k 1 j 1
s.t.
P Pmax
Pmin Generator Constraints
D is an MxM matrix that will have its diagonal as the admittances of the lines.
M = number of branches
A is an M x (N-1) node-arc incidence matrix that describe the flow from bus i to
bus j.
N= number of busses
B’ is the DC power flow matrix of dimension (N-1)x(N-1), where N is the
number of buses in the network, obtained as follows:
1. Replace diagonal element B’kk with the sum of the non-diagonal elements
in row k. Alternatively, subtract bk (the shunt term) from Bkk, and multiply
by -1.
2. Multiply all off-diagonals by -1.
θ is the vector of nodal phase angles for buses 2,…N
PD is the Power demanded at each bus.
Aeq matrix – The number of columns is the sum of the number of generators,
lines, and buses. The number of rows is the sum of the number of
lines and buses.
Module PE.PAS.U22.5 Composite system reliability 21
Example:
This is the basic system that will be used for the example.
Pg2
Pg1
1 2
y12 =-j10
y14 =-j10
y13 =-j10 Pd2=1pu y23 =-j10
y34 =-j10
4 3
Pg4 Pd3=1.1787pu
This graph gives the basic representation of the cost curves division for a generator
Ci
$/hr
Pi3
Pi1 Pi2
Pi,min Pi (MW)
We will use linearized representation, but to begin with, we assume quadratic
representation for the three cost curves of the example.
Break points for PW linear approximation of cost curves given as MW, Cost
The slope is then computed for each of the segments (1-2), (2-3), and (3-4) for each
curve. This is done by the following equation:
This process is done for each section and each generator. The slopes of the piecewise
linear approximations of the cost curves are shown below:
K1 ( P11 , P11 , P11 ) C1 ( P1,min ) s11 P11 s12 P12 s13 P13
K 2 ( P21 , P22 , P23 ) C 2 ( P2,min ) s 21 P21 s 22 P22 s 23 P23
K 3 ( P41 , P41 , P43 ) C3 ( P3,min ) s 41 P41 s 42 P42 s 43 P43
The limits on generation for each section must be calculated. This is given by:
This process is done for each section of the curve for each generator. Representing the
vector of piecewise linear generation values as P’, we have:
Module PE.PAS.U22.5 Composite system reliability 23
P11 0 0.50
P 0 0.60
12 Unit 1
P13 0 0.40
P21 0 0.325
P P22 Unit 2 0 P 0.60
P23 0 0.20
P 0 0.45
41
P42 Unit 4 0 0.50
0 0.40
P43
Next, the T matrix must be built. This is a matrix that operates on the piecewise linear
generation vector P’ to obtain generation injections, i.e., it sums the generation variables
for each unit to obtain the total generation at that unit. For this case, T will be 4 rows
high, and 9 columns long.
1 1 1 0 0 0 0 0 0
0 0 0 1 1 1 0 0 0
T
0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 1 1 1
P11
P
12
P13
P1 1 1 1 0 0 0 0 0 0
P 0
P21
2 0 0 0 1 1 1 0 0
P22
P3 0 0 0 0 0 0 0 0 0
P23
P4 0 0 0 0 0 0 1 1 1
P
# of buses 41
P42
P43
The next step is to get the respective parts for the DC Power Flow
( B PD T P Pmin ). We first obtain the B’ matrix.
Module PE.PAS.U22.5 Composite system reliability 24
From the above, we obtain the B’ matrix from the Y-bus, as follows:
5. Remove the “j” from the Y-bus.
6. Replace diagonal element B’kk with the sum of the non-diagonal elements in row k.
Alternatively, subtract bk (the shunt term) from Bkk, and multiply by -1.
7. Multiply all off-diagonals by -1.
Comparison of the numerical values of the Y-bus with the numerical values of the B’
matrix for our example will confirm the above procedure:
30 10 10 10 30 10 10 10
10 20 10 0 10 20 10
0
Y j B'
10 10 30 10 10 10 30 10
10 0 10 20 10 0 10 20
Another way to remember the B’ matrix is to observe that since its non-diagonal elements
are the negative of the Y-bus matrix, the B’ non-diagonal elements are susceptances..
P D1 0
PD is the vector of load values, in per unit: P P D 2 1 p.u.
D
P D 3 1.1787
P D4 0
Pmin is the minimum amount of power that each generator can produce.
Pmax is the maximum amount of power that each generator can produce.
Module PE.PAS.U22.5 Composite system reliability 25
0.50 Pg 1 0.200
0.375 P 0.150
g2
0 Pg 3 0
P
0.45
g4 0.180
Pmin Pmax
We account for the lower limit in the DC power flow equation through Pmin.
Expressing the DC Power Flow Equation ( B PD T P Pmin ).
The D-matrix is formed by placing the admittances of each branch along the diagonal of
an M x M matrix, where M=5.
10 0 0 0 0
0 10 0 0 0
D 0 0 10 0 0
0 0 0 10 0
0 0 0 0 10
The node-arc incidence matrix is found by first defining a direction of power flow in the
system.
1 2
2
1
5 3
4
4 3
A branch is said to “begin” at node j if the power flowing across branch k is defined
positive for a direction from node j to the other node.
Module PE.PAS.U22.5 Composite system reliability 26
A branch is said to “terminate” at node j if the power flowing across branch k is defined
positive for a direction to node j from the other node.
1 0 0- 1
1 -1 0 0
A 0 1 1 0 (Has four columns because there are four bus angles)
0 0 1 1
1 0 1 0
To begin with, we will represent unconstrained transmission using very large limits.
500 500
500 500
500 P B 500 per-unit.
500 500
500 500
In this example, we have 4 buses, 5 branches, and 9 generation sections. So that sums up
to 18, so that is the number of columns in the Aeq matrix. For the number of rows, we
have 5 branches and 4 buses, which sums up to 9 and is the number of rows in this
matrix.
The last 4 columns on the first 5 rows is the DxA matrix.
The first 14 columns on the first 5 rows are the elements in the line flow equations
that multiply the variables P11, P12, P13, P21, P22, P23, P41, P42, P43, PB1, PB2, PB3, PB4,
and PB5. P11, P12, P13, P21, P22, P23, P41, P42, P43 are set to zero because they are the
Module PE.PAS.U22.5 Composite system reliability 27
generations not the line flow. PB1, PB2, PB3, PB4, and PB5 are set to -1 to say that there
is a branch in a specific location.
The first 9 columns of the last 4 rows multiply with the generation variables P11, P12,
P13, P21, P22, P23, P41, P42, P43,. The demand (PD) will be given to us in the b matrix
and Pgk-Pdk is in the b matrix.
The 10-14 columns of the last 4 rows correspond to the line flow variables, but these
are zeros because they are not the DC power flow equation.
The last 4 columns of the last 4 rows correspond to the DC power flows that include
all the angles.
We can see that these elements will occupy the upper right hand corner of Aeq. So that
will take care of the last 4 columns in the first 5 rows.
What about the first 14 columns? These elements are the equations that are multiplied by
the variables P11, P12, P13, P21, P22, P23, P41, P42, P43, PB1, PB2, PB3, PB4, and PB5. Since we
do not use the generation variables within the line flow equations, the first 9 columns of
these top 5 rows will be zeros. The last 5 columns in these top 5 rows will also be zeros,
except the one element in each of these rows that multiply the corresponding line flow
variable, and that element will be -1.
P11
P
12
P13
P
21
0 P22
0 0 0 0 0 0 0 0 1 0 0 0 0 10 0 0 10 0
P23
0 0 0 0 0 0 0 0 0 0 1 0 0 0 10 10 0 0 0
0 P
0 0 0 0 0 0 0 0 0 0 1 0 0 0 10 10 0 41 0
P
0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 10 10 42 0
P
Aeq x 0 0 0 0 0 0 0 0 0 0 0 0 0 1 10 0 10 0 P43 0
B1
P 0
B2
P 1
B3
P 1.1787
B4
P 0
B5
1
2
3
4
The B’ matrix deal with the bus angles. This is inserted into the last 4 columns and the
bottom 4 rows. The resulting matrix appears as:
Module PE.PAS.U22.5 Composite system reliability 28
P11
P
12
P13
P
21
0 P22
0 0 0 0 0 0 0 0 1 0 0 0 0 10 0 0 10 0
P23
0 0 0 0 0 0 0 0 0 0 1 0 0 0 10 10 0 0 0
0 P
0 0 0 0 0 0 0 0 0 0 1 0 0 0 10 10 0 41 0
P
0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 10 10 42 0
P43
Aeq x 0 0 0 0 0 0 0 0 0 0 0 0 0 1 10 0 10 0 P 0
B1
30 10 10 10 P 0
B2
10 20 10 0 P 1
B3
10 10 30 10 P 1.1787
B4
10 0 10 20 P 0
B5
1
2
3
4
Once again, we need to consider the first eight columns. Columns 4-8 correspond to the
line flow variables, which do not appear in the DC power flow equations, so these will be
zero.
Given that the demand for Generator 2 = 1.0 per unit and Generator 3 = 1.178 per unit
Pd1=0, Pd2=1.0, Pd3=1.1787, Pd4=0, Pg3=0
Since these are constants, they can go to the right-hand side. This means on the left-hand
side, the injection, which is defined as Pgk-PDk should be negative. That is, we should see
on the left-hand-side –Pgk+PDk. But now we will take the load term onto the right-hand-
side by subtracting it from both sides.
Thus, we see that the load term should show up on the right-hand-side as a negative
number
Module PE.PAS.U22.5 Composite system reliability 29
P11
P
12
P13
P
21
0 0 0 0 0 0 0 0 0 1 0 0 0 0 10 P22
0 0 10 0
P23
0 0 0 0 0 0 0 0 0 0 1 0 0 0 10 10 0 0 0
0 0 0 0 0 0 0 0 0 0 0 1 0 0 P
0 10 10 0 41 0
P
0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 10 10 42 0
P43
Aeq x 0 0 0 0 0 0 0 0 0 0 0 0 0 1 10 0 10 0 P 0
B1
1 1 1 0 0 0 0 0 0 0 0 0 0 0 30 10 10 10 P 0
B2
0 0 0 1 1 1 0 0 0 0 0 0 0 0 10 20 10 0 P 1
0 0 0 0 0 0 0 0 0 0 0 0 0 0 10 10 B3
30 10 P 1.1787
B4
0 0 0 0 0 0 1 1 1 0 0 0 0 0 10 0 10 20 P 0
B5
1
2
3
4
Putting all of the constraints together, and converting the generator constraints to per
unit, the following vector of constraints is given.
0 P11 .5
0 .6
P12
0 P13 .4
0 P21 .325
0 P22 .6
0 P23 .2
0 P .45
41
0 P42 .5
0 P .4
43
5 PB1 5
P
5 B 2 5
5 PB 3 5
P
5 B 4 5
5 PB 5 5
1
2
3
4
P11 0.5000
P 0.0000
12
P13 0.0000
P 0.3250
21
22
P 0.6000
P 0.0000
23
P41 0.4500
P42 0.3037
P 0.0000
x P43
B1 - 0.1670
PB 2 0.2473
PB 3 0.1723
PB 4 0.5867
P B 5
0.4197
1 0.0126
2 - 0.0122
3 - 0.0294
4 0.0293
C1 0.00533 (50 ) 2 11 .669 (50 ) 213 .1 12 .46 (50 ) 186 .88
C 2 0.00889 (95 .5) 2 10 .333 (92 .5) 200 11 .29 (32 .5) 12 .11(60 ) 143 .36
C 4 0.00741 (75.37 ) 2 10.833 (75.37 ) 240 11.83(45) 12 .54(30.37 ) 185 .39
0.5000
0.0000
0.0000
PG1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0.5 PG1
P 0.3250
G 2 0 0 0 1 1 1 0 0 0 0 0 0 0 0 0.6000 0.925 PG 2
PG 3 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 PG 3
0.0000
PG 4 0 0 0 0 0 0 1 1 1 0 0 0 0 0 0.4500 0.7537 PG 4
P 0 0 0 0 0 0 0 0 0 1 0 0 0 0 - 0.16701 PB1
B1 0.3037
PB 2 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0.24734 PB 2
P 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0.0000 0.17234 P
B3 - 0.1670 B3
PB 4 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0.58669 PB 4
0.2473
PB 5
0 0 0 0 0 0 0 0 0 0 0 0 0 1
0.41967 PB 5
0.1723
T
0.5867
0.4197
Part of x
0 P11 .5
0 .6
P12
0 P13 .4
0 P21 .325
0 P22 .6
0 P23 .2
0 P .45
41
0 P42 .5
0 P .4
43
5 PB1 5
P
0.15 B 2 0.15
5 PB 3 5
P
5 B 4 5
5 PB 5 5
1
2
3
4
P11 0.5
P 0
12
P13 0
P 0.325
21
P22 0.6
P 0.1946
23
P41 0.45
P42 0.1090
P 0
x P43
B 1 - 0.0697
PB 2 0.15
PB 3 0.26967
PB 4 0.4894
PB 5 0.4197
1 0.0125
2 - 0.0025
3 - 0.0295
4 0.0195
C1 0.00533 (50 ) 2 11 .669 (50 ) 213 .1 12 .46 (50 ) 186 .88
C 2 0.00889 (111 .96 ) 2 10 .333 (111 .96 ) 200 11 .29 (32 .5) 12 .11(60 ) 12 .82 (10 .9) 235 .06
C 4 0.00741 (55.9) 2 10.833(55.9) 240 11.83(45) 12.54(10.9) 199 .6835
The most noticeable change in the system is how the generation shifted at each
generator to account for the constraint on Bus 2.
Module PE.PAS.U22.5 Composite system reliability 36
0.5
0
0
PG1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0.5 PG1
P 0.325 P
G 2 0 0 0 1 1 1 0 0 0 0 0 0 0 0 0.6 1.1197 G2
PG 3 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 PG 3
0.1946
PG 4 0 0 0 0 0 0 1 1 1 0 0 0 0 0 0.45 0.55903 PG 4
P 0 0 0 0 0 0 0 0 0 1 0 0 0 0 - 0.069675 PB1
B1 0.1090
PB 2 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0.15 PB 2
P 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0.26967 PB 3
B3 - 0.0697
PB 4 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0.48935 PB 4
0.15 0.41967 P
0 0 0 0 0 0 0 0 0 0 0 0 0 1
PB 5 0.26967 B5
T
0.4894
0.4197
Part of x
0.5 0.5 0
0.925 1.1197 - 0.1947
0 0 0
0.7537 0.55903 0.19467
P - 0.16701 - 0.069675 - 0.0974
0.24734 0.15 0.09734
0.17234 0.26967 - 0.0973
0.58669 0.48935 0.09734
0.41967 0.41967 0
Poriginal Pconstrained
A equivalent matrix
RHS equality constraints
Lower bounds
Upper bounds
We find the objective function, lower bounds, and the upper bounds from the cost curve
data. The A matrix, RHS of inequality constraints, RHS of equality constraints, and the
A equivalent matrix are found using MatLab code using data from PSSE.
Module PE.PAS.U22.5 Composite system reliability 38
We can obtain the flows on all circuits given the injections at all
buses, as shown below:
P B' (1)
P B ( D A) (2)
Inverting eq (1) yields:
1
B' P (3)
Substitution of (3) into (2) yields:
P B ( D A)B' P
1
(4)
Bus this is not what we want. What we want is the change in flow
on circuit b given a change in injection at bus k.
Pk Pk Pk (5)
0
PN PN PN
The change in circuit flows can then be expressed as
Module PE.PAS.U22.5 Composite system reliability 39
PB PB P
0
B
( D A)B ' P P
1
0
(6)
( D A)B ' P
1
Now let the ∆P vector be all zeros except for the element
corresponding to the kth bus, and assign this bus an injection
change of 1.
P2 P2 0
P P 0
3 3
P
P P
k k 1 (7)
PN PN 0
Then
PB1 t1, k 0
P t 0
B 2 2 , k
1
P B ( D A)B '
PBb t b , k 1 Pk (8)
BM M , k
P t 0
Question: Does the above equation imply the injection is changed
at only one bus? Answer: No. There is an equal and opposite
change to the injection at the swing bus, in this case, bus 1.
PBb
tb, k
Pk Reallocation
Policy
This is denoted as ali in the W&W (ref [1]) text (see eq. 11.1).
Example 1:
We consider an example used in Appendix 1, illustrated below.
Compute the generation shift factors for all branches
corresponding to an increase in bus 2 injection and a decrease in
bus 3 injection.
Pg2=2pu
Pg1=2pu
1 2
y12 =-j10
y14 =-j10
y13 =-j10 Pd2=1pu y23 =-j10
y34 =-j10
4 3
Pg4=1pu Pd3=4pu
t1, 23 10 0 0 00 0 0 - 1
t
10 0 0 0 - 1 0 0 20 10 0 1
1
2 , 23 0
t 3,23 0 0 10 0 0 1 - 1 0 10 30 10 1
t
4, 23 0 0 0 10 0 0 - 1 1 0 10 20 0
t 5, 23 0 0 0 0 10 0 - 1 0
0 0 10
10 0 0 0.0625 0.025 0.0125 1
10 10 0 0.025 0.05 0.025 1
0 10 10 0.0125 0.025 0.0625 0
0 10 0
0 0 10 0.125
10 0 0 0.0375 0.375
10 10 0 0.025 0.625
0 10 10 0.0125 0.125
0 10 0 0.25
Note that the above generation shift factors are for a “double shift.”
You can think of it like this. A generation shift factor for branch b,
bus k would be tb,k and another generation shift factor for branch b,
bus j would be tb,j. If we have an injection increase at bus k of ∆Pk
and an injection increase at bus j of ∆Pj (negative), then
Pb tb, k Pk tb, j Pj (9)
Pg2=2pu
Pg1=2pu
1 2
2
1
5 Pd2=1pu 3
4
4 3
Pg4=1pu Pd3=4pu
2 3 4
20 10 0 2
B ' 10 30 10 3
0 10 20 4
Recalling that all branch admittances of our network are –j10, what
would these four elements be if branch #3 (between buses 2 and 3)
were not there?
2 3 4
10 0 0 2
B ' out 0 20 10 3
0 10 20 4
What is the difference between B’ and B’out?
10 0 0 20 10 0 10 10 0
B' B' out B' 0 20 10 10 30 10 10 10 0
0 10 20 0 10 20 0 0 0
Notice that the elements in ∆B’ are all multiples of B’23=-10, i.e.,
1 1 0
B' 10 1 1 0
0 0 0
The above matrix can be expressed as
1 1 0 1
1 1 0 11 1 0
0 0 0 0
From this simple illustration, we can see a generalization, that
whenever we remove a branch between buses i and j, with
corresponding B’ matrix element B’ij, the B’ matrix will change as
indicated below.
Module PE.PAS.U22.5 Composite system reliability 45
Inconsistency: In Section i j
6.0, we used (i,j) to
indicate terminals of the
circuit to be loaded (l) 0 0 0 0
and (m,n) to indicate
terminals of the circuit to
be outaged (k). In the
development of this 0 1 1 0 i
section, the nomenclature B ' B ' out B ' bij
on terminal number has
been reversed, i.e., (i,j) 0 1 1 0 j
becomes the terminals of
the circuit to be outaged
(k) and (m,n) becomes 0 0 0
the terminals of the
0 (22)
where
circuit to be loaded (l). bij
is the susceptance of branch i-j (and is a negative number
for any standard circuit, which is inductive). We use bij instead of
-B’ij in order to ensure we have a defined term even when i or j are
the swing bus.
0
1 i
e ij
1 j (24)
0
then (23) becomes
P B'out ( ) (28)
(We are assessing the effects of only a lone outage, so the injection
vector P does not change). Substituting (27) into (28), we obtain
T 1
B'bij eij eij P
(30)
We do not seem to have made much progress, because we still
have to take an inverse…
B'cd T 1 1 1
B' B' c I
(M )
T 1 1 T 1
d B' c d B'
where I(M) is the M×M identity matrix.
It may not be very obvious at this point that MIL will help us,
since we see 4 different inverses on the right-hand-side of MIL.
Let’s apply MIL to the inverted term of (30) to see what happens.
B'1 B'1 bij e ij I ( M ) eTij B'1 bij e ij 1 T
e ij B'1
(32)
Pulling out the scalar multiplier bij from where it appears in both
the numerator and denominator, we have
bij B ' 1 e ij eTij B ' 1
T 1
B 'bij e ij e ij B ' 1
1 bij eTij B '1 e ij (34)
1 1
T
T 1 B ' e e B '
B'1
ij ij
B'bij e ij e ij
1
eTij B'1 e ij (35)
bij
What we have just expressed in (35) is the inverted term on the
right-hand-side of (30), repeated below for convenience:
T 1
B'bij eij eij P (30)
Substituting (35) into (30), we obtain:
1 T 1
1 B ' eij eij B '
B ' P
1
eTij B '1 eij (36)
b
ij
Distributing the injection vector P results in
1 T 1
1 B ' e e
ij ij B ' P
B' P
1
eTij B'1 eij (37)
bij
But θ=B’-1P, and therefore we can replace the corresponding
expressions in both right-hand-side terms to obtain:
B'1 eij eTij
1
eTij B'1 eij (38)
bij
We can simplify a little more by investigating eijTθ in the
numerator. This would be:
Module PE.PAS.U22.5 Composite system reliability 50
2
i
eTij 0 1 1 0 i j
i j j (39)
n 1
Substituting (39) into (38) results in:
1
B' eij ( i j )
1
eTij B'1 eij (40)
bij
Now we have only two inverses left. Interestingly, they both pre-
multiply eij. That is, we observe that both inverses appear in
B’-1eij, an n×1 vector.
g ij
2
gi
ij
eTij g ij 0 1 1 0 g iij g ijj
i j g ijj (44)
ij
g n 1
Therefore, (43) becomes
( i j )
g ij
1
( giij g ijj ) (45)
bij
and
Module PE.PAS.U22.5 Composite system reliability 53
eTmn g ij g m
ij
g nij (52)
Substituting (51) and (52) into (50) results in
ij
fk 0 ( gm g nij )
f l bmn
1 bij ( giij g ijj ) (53)
Example 5:
Consider our 4-bus, 5-branch example problem again. Compute gij
for a line 2-3 outage. Then use it to compute the post-contingency
flow on circuit 3-4.
Solution:
Recall (42):
B' g eij
ij
20 10 0
B 10 30 10
0 10 20
and e23 is given by
1 i 2
e 23 1 j 3
0
And so our equation is:
20 10 0 g 2 1
23
10 30 10 g 23 1
3
0 10 20 g 423 0
Performing LU decomposition, we obtain
20 0 0 1 0.5 0
L 10 25 0
U 0 1 0.4
0 10 16 0 0 1
Notice that the above factors need be computed only one time;
they may subsequently be applied to obtain the g-vector for outage
of any circuit. In this case, we are interested in outage of the line
from bus 2 to bus 3, therefore we write
20 0 0 w1 1 w1 0.05
L w e23 10 25 0 w2 1 w2 0.02
0 10 16 w3 0 w3 0.0125
1 0.5 0 g 223 0.05 g 223 0.0375
U g 23 w 0 1 0.4 g323 0.02 g323 0.025
0 0 1 g 423 0.0125 g 423 0.0125
Then we can compute the LODF for the circuit 3-4 after outage of
circuit 2-3. First, however, note that the direction of positive flow
on circuit 3-4 is defined by the node-arc incidence matrix as being
from bus 4 to bus 3 (see below right-hand figure).
Module PE.PAS.U22.5 Composite system reliability 55
Pg2=2pu Pg2=2pu
Pg1=2pu Pg1=2pu
1 2 1 2
y12 =-j10 2
y14 =-j10 1
y13 =-j10 Pd2=1pu y23 =-j10 5 Pd2=1pu 3
y34 =-j10 4
4 3 4 3
P B ( D A)B' P
1
PB1 10 0 0 0 0 0 0 - 1
P 0 10 0 0 0 - 1 0 0 20 10 0 -1 1
B2
PB 3 0 0 10 0 0 1 - 1 0 10 30 10 4
PB 4 0 0 0 10 0 0 - 1 1 0 10 20 1
PB 5 0 0 0 0 10 0 - 1 0
PB1 0.25
P 0.25
B2
PB 3 1.25
P
B4 1 . 25
PB 5 1.5
Observe that the flow on circuit 3-4 is 1.25 pu.
With circuit 2-3 out, we obtain:
PB1
P 10 0 0 0 0 0 1 10 0 -1
B 2 0 10 0 0 1 0 0 0 1
PB 3 0 20 10 4
0 0 10 0 0 1 1
PB 4 0 0 0 10 0 1 0 0 10 20 1
PB 5
Pg2=2pu Pg2=2pu
Pg1=2pu Pg1=2pu
1 2 1 2
y12 =-j10 2
y14 =-j10 1
y13 =-j10 Pd2=1pu y23 =-j10 5 Pd2=1pu 3
y34 =-j10 4
4 3 4 3
PB1 0.6667
P 1.0
B3
PB 4 1.6667
B5
P 2. 3333
Therefore the change in flows due to outage of circuit 2 (connected
between buses 2 and 3) is
0.6667 0.25 0.4167
0 0.25 0.25
Pb Pb P b 1.0 1.25 2.25
out
1 . 6667 1. 25 0. 4167
2.3333 1.5 0.8333
where we observe that ΔPb4=0.4167, as indicated in the calculation
of Δfl on pg. 33.
…Also should use W&W’s method here.
Example 6:
For outage of branch connected to swing.
Do it three ways as in Example 6:
a. Use the above “fast” method
b. Use DC flow with and without outage;
c. Use W&W’s method.
References
[1] A. Wood and B. Wollenberg, “Power generation, operation, and control,” Second edition, John Wiley & Sons, 1996.
[2] “Transmission system reliability methods,” project 1530-1, prepared by Power Technologies Incorporated (PTI),
EPRI report EL-2526, 1982.
[3] Feng Xia; Meliopoulos, A.P.S, “A methodology for probabilistic simultaneous transfer capability analysis”.;
Power Systems, IEEE Transactions on, Volume 11, Issue 3, Aug. 1996 Page(s):1269 – 1278.
[4] N. Dag Reppen, presentation slides given at EPRI Workshop on Reliability Analysis, San Diego, California, Feb 27,
2001.
Module PE.PAS.U22.5 Composite system reliability 58
[5] R. A. Leon Candela, “Sensor network design for a secure electric energy infrastructure,” M.S., thesis, Iowa State
University, April 2005.
[6] Q. Chen and J. McCalley, “Identifying High-Risk N-k Contingencies for On-line Security Assessment,” IEEE
Transactions on Power Systems, 2005.
[i] A. Debs, “Modern Power Systems Control and Operation,” Kluwer, 1988.
[ii] A. Monticelli, “State estimation in electric power systems, a generalized
approach,” Kluwer, 1999.