Basic Probability and Reliability

Concepts

Roy Billinton
Power System Research Group
University of Saskatchewan
CANADA

14
Basic Probability

Probability
- measure of chance
- quantitative statement about the
likelihood of an event or events

0 0.5 1.0
Absolute Toss of a Absolute
impossibility fair coin certainty

15
Basic Probability
Apriori Probability
Number of S uccesses
P[success] 
Number of Possible Outcomes
Number of Failures
P[Failure] 
Number of Possible Outcomes
1
Coin - P[Head] 
2
1
Die - P[S ix]
6 16
Basic Probability
Consider two dice – what is the probability of
getting a total of 6 in a
single roll?

Possible outcomes = 6×6 = 36 ways

Successful outcomes = (1+5) (2+4) (3+3) (4+2) (5+1)
= 5 ways
P [Six] = 5/36
Total 2 3 4 5 6 7 8 9 10 11 12
Prob. in 36ths
17
Basic Probability
Relative frequency interpretation of probability
f
P[of a particular event occuring]  lim
n  n

n  number of times an experiment is repeated

f  number of occurrence s of a particular outcome.
Consider tossing a coin, rolling a die.
Estimate the unavailability or probability of finding a piece
of equipment on outage at some distant time in the future.

Unavailabi lity 
 (Outage Time)
 (Outage Time)   (Operating Time)
18
Basic Probability
Basic Rules
1. Independent events: Two events are said to be
independent if the occurrence of one event does not
affect the probability of occurrence of the other
event.
2. Mutually exclusive events: Two events are said to
be mutually exclusive or disjoint if they cannot both
happen at the same time.
3. Complimentary events: Two outcomes of an event
are said to be complimentary if, when one outcome
occurs, the other cannot occur.
19
Basic Probability
4. Conditional events: Conditional events are events
which occur conditionally on the occurrence of another
event or events.
Consider two events A and B and consider the probability of
event A occurring under the condition that B has occurred.
This probability is P(A|B).
Number of ways A and B can occur
P(A | B) 
Number of ways B can occur
S AB
P(A  B) 
A B S
B
P(B) 
S
S  P(A  B) P(A  B)
P(A | B)  
S  P(B) P(B) 20
Basic Probability
Independent events

P(A | B)  P(A)
P(A  B)  P(A | B)  P(B)
 P(A)  P(B)

n
P(A1  A2  A3......n)   P(A i )
i 1

21
Basic Probability
The occurrence of at least one of two events A and
B is the occurrence of A OR B OR BOTH.

P(A  B)  P(A)  P(B) - P(A  B)

 P(A)  P(B) - P(A | B)  P(B)
S  P(A)  P(B)  P(A)  P(B)
if A and B are independent events
A B

22
Basic Probability
P(A  B)  P(A | B)  P(B)
B i  mutually exclusive events
P(A  B1 )  P(A | B1 )  P(B 1 )
B1 P(A  B 2 )  P(A | B 2 )  P(B 2 )
B2 A P(A  B 3 )  P(A | B 3 )  P(B 3 )
P(A  B 4 )  P(A | B 4 )  P(B 4 )
B4
B3 4 4

 P(A  B )   P(A | B )  P(B )

i 1
i
i 1
i i

n
P(A)   P(A | B i )  P(B i )
i 1
23
Expectation
n
Discrete distribution E   xipi
i 1

Continuous distribution
E   x  f(x)dx
0
Example:
Prize  $10.00
1
P(Winning) 
5
1 4
Expectation   10   0  $2.00
5 5
24
Example
Probability that a 30 year old man will survive
a fixed time period is 0.995. Insurance company offers
a $2000 policy for $20. What is the company’s
expected gain?

Probability Gain
0.995 20
0.005 -1980

E (Gain)= 0.995∙(20)+0.005 ∙(-1980)

= $10.00
25
 Expectation Example
The distribution (discrete) of the power output from a 100 MW
wind farm is given in the table below.
What is the expected power output?

i Capacity Probability xi.pi

(xi MW ) (pi) (MW)
1 100 0.03 3.00
2 75 0.08 5.25
3 50 0.15 7.50
4 25 0.35 8.75
5 0 0.39 0.00
Expected Power Output (MW) = 25.25

26
 Expectation

n
E x p
i 1
i i


E 
-
x  f(x)dx

27
 Mean Time to Failure

E(t) =  t.f(t)dt l
0
λt

f(t) = l e
Q(t)
MTTF =  t.f(t)dt

f(t)
0 R(t)
0.37l

1
  t . λ e-λtdt 
0 λ
0 t 1/l
time

Expectation Indices
• Expected Frequency of Failure
• Expected Duration of Failure
• Expected Annual Outage Time
• Expected Energy Not Supplied
• Expected Annual Outage Cost 28
 Binomial Distribution
(p+q)2 = p2 + 2pq + q2 (p+q)3 = p3 + 3p2q + 3pq2 + q3

General Expression for Binomial Distribution:

(p + q)n = pn + n pn-1 q +…. n(n  1)..[n - (r - 1)] pn-r qr + …. + qn

r!
n = number of components or trials n!
p = probability of success r! (n - r)! = nCr
q = probability of failure
Probability of exactly r failures (and n-r successes),
Pr = nCr p(n-r) qr
AIEE Committee Report, Tables of Binomial Probability Distribution to Six
Decimal Places, AIEE Transactions (August 1952), pp. 597-620.
29
Binomial Distribution
n r n! r n r
Pr  n C r p q
r
 p q
r! (n  r)!
Consider a 3*5 MW unit plant. Each unit has a F.O.R
of 3%.
(R  Q)3  R 3  3R 2 Q  3RQ 2  Q 3
Capacity Capacity
Units Out Probability
Out (MW) Available (MW)
0 0 15 0.912673
1 5 10 0.084681
2 10 5 0.002619
3 15 0 0.000027
1.000000 30
Boiler Circulating Pumps
3 pumps – each pump rated at 90% F.L.R
pump unavailability = 0.01

Pumps Unit Capacity

Probability Expectation
Out Out
0 - 0.97029890 -
1 - 0.02940299 -
2 10% 0.00029700 0.00297
3 100% 0.00000100 0.00010
0.00307

31
 3 Pump Systems
Pump Rating Expected % Capacity Loss

100 0.00010
90 0.00307
80 0.00604
70 0.00901
60 0.01198
50 0.01495
40 0.60598
32
Basic Reliability
Let R=P [Success]
Q=P [Failure]
R+Q=1
Series Systems 1 2

Qs  1  R S Rs  R1  R 2
 1  R1  R 2 n
 Ri
 1 - (1 - Q 1 )(1  Q 2 ) i 1

 Q1  Q 2  Q1  Q 2
33
 Series System
If each component has
a reliability of 0.9. 1.0 0.999
0.9
Number of Reliability
Components 0.8

System Reliability
0.7
1 0.9
2 0.81 0.6
0.98
3 0.729 0.5
4 0.6561 0.4
5 0.59049 0.3
10 0.348678
0.2
20 0.121577 0.9
50 0.005154 0.1
0.0
0 10 20 30 40 50 60 70
Number of Components

System Reliability decreases as the number of components increases in a Series

System. The number on the curve is the reliability of each component. 34
Basic Reliability

Parallel Redundant Systems

Q s  Q1  Q 2
1
R s  1  QS
 1  Q1  Q 2
2
 1 - (1 - R 1 )(1  R 2 )
 R1  R 2  R1  R 2
35
Parallel System

Number of Components Reliability

1 0.9
2 0.99
3 0.999
4 0.9999
5 0.99999

36
Basic Reliability
Series/Parallel Systems

2 3
1
4
Redundant

Rs = R1[R 2 R 3  R 4  R 2 R 3 R 4 ]
37
 Binomial Systems

m /n System

Identical
Components

(R  Q)5  R 5  5R 4 Q  10R 3 Q 2  10R 2Q 3  5RQ 4  Q 5

Rs Qs
System Criterion = 3/5
38
Conditional Probability Approach
If the occurrence of an event A is dependent upon
a number of events Bj which are mutually exclusive.
j
P(A)   P(A | B i )  P(B i )
i 1

If A is defined as system success

P(SystemSuccess) P(SS | B X )  P(B X )  P(SS | By )  P(B y )

If A is defined as system failure

P(SystemFailure)  P(SF | B X )  P(B X )  P(SF | By )  P(B y )

39
 Series System

1 2

P(S S )  P(S S | 1 is good)  R 1  P(S S | 1 is bad)  Q1

 R 1R 2 0  Q1
 R1R 2

40
 Parallel System

P(SS)  P(SS | 1 is good)  R 1  P(SS | 1 is bad)  Q1

 1 R 1  R 2  Q1
 R1  R 2  R1R 2
41
 Non Series/Parallel Systems

1 2

Output
3 4

P(SS)  P(SS | 1 is good)  R1  P(SS | 1 is bad)  Q1

 [R 2 R 4 - R 2  R 4 ]  R1  R 3  R 4  Q1

42
Minimal Cut Set Method
Cut Set – A set of components which if removed
from the network separate the input from the output.
i.e. cause the network to fail.
Minimal Cut Set – Any cut set which does not
contain any other cut sets as subsets.
P{S ystemFailure}  P{Union of All Cut S ets}
 P{Union of All Minimal Cut S ets}
  P{Min Cut S ets}

This is a good approximation for highly reliable

components.
43
P{S ystemFailure} P{Union of All Minimal Cut S ets}
 P{C 1  C 2  C 3 .....  C n }
n
  P(C i )
i 1

Consider
1 2

Q S  P{C 1  C 2 }
 P(C 1 )  P(C 2 )  P(C 1  C 2 )
 Q1  Q 2  Q1  Q 2
 Q1  Q 2
44
Basic Reliability
Consider: 1 2

3
Cuts Min Cuts Probability
1,3 1,3 Q1Q3
2,3 2,3 Q2Q3
1,2,3 --- -
Qs<Q1Q3+Q2Q3
Complete Equation:
Q S  Q 3 [Q1  Q 2  Q 1Q 2 ]
 Q 1Q 3  Q 2 Q 3  Q 1Q 2 Q 3
45
Mission Orientated Systems
Reliability is the probability of a device or system
performing its purpose adequately for
the period of time intended under the
operating conditions encountered.

System λ System
Up Down

R(t)  e  λt
Where l  component failure rate

46
 Mission Reliability

 λt
Q(t) f(t) = l e
f(t)

R(t)
0.37l

0 t 1/l time

47
Conventional Bathtub Curve

Region 1 Region 2 Region 3

Hazard rate

De- Normal operating Wear out

Bugging Or useful life
Operating Life
Typical Electric Component Hazard Rate as a Function of Age
48
 Network Models and Mission Reliability

Series Systems Parallel Systems

1
1 2
2

product rule of reliability product rule of unreliability

R s  R1R 2
 e  λ 1t  e  λ 2 t R s  R1  R 2  R1R 2
 e -(λ 1  λ 2 )t R S  e λ1t  e λ 2t  e -(λ 1  λ 2 )t
- λ i t
e 49
Basic Reliability
Mission systems

*Develop the basic equations

*Substitute
 λt
R(t)  e
 λt
Q(t)  1  e

50
 System Reliability and Availability

Reliability –
probability of a system staying in the operating state without failure

1 l 2
Up Down
R(t) = e λt

Availability –
probability of finding a system in the operating state at some time
into the future

l μ λ
1 2 A(t) = + e-(l + )t
Up Down λ μ λ μ


51
 System Reliability and Availability

μ λ
A(t) = + e-(l + )t
λ μ λ μ

R(t) = e λt

In the limiting state:

μ λ
A= U=
λ μ λ μ

52
 Markov Analysis

Application:
Random behaviour of systems that vary discretely or
continuously with respect to time and space.

Reliability Evaluation:
Space: Normally discrete and identifiable states.
Time: Discrete (Markov Chain)
Continuous (Markov Process)

53
 Markov Analysis
Applicability
Systems characterized by a lack of memory. Future
states are independent of all past states except the
immediately proceeding one.
System process must be stationary. Probability of
making a transition from one state to another is the
same (stationary) at all times. The state probability
distribution is characterized by a constant transition
rate.

54
 Markov Analysis
State Space / State Transition Diagram
1 λ 2

Up Down
µ
Stochastic transitional probability matrix P

1 1 – λΔt λΔt
P =
2 µΔt 1- µΔt

State probabilities after n increments = Pn

55
 Markov Analysis
Limiting state probability vector = [ P1 P2 ]

[ P1 P2 ] P = [ P1 P2 ]

[ P1 P2 ] 1 – λΔt λΔt = [ P1 P2 ]
µΔt 1 - µΔt

P1 = µ
-λP1 + µP2 = 0
λ+µ
λP1 - µP2 = 0
P1 + P2 = 1.0 P2 = λ
λ+µ
56
 System Availability

Up
1
Up
Down
0 Time
l 

2
Down Up MTTF
MTBF MTTR
Down
0 Time

total up time 1
MTTF = = MTBF = 1/F
# of failures l

total down time 1

MTTR = average repair time = r = =
# of failures 
57
 System Availability Example

Example: If the failure rate of a system is 1.5 failures/year and

the average repair time is 10 hours, what is the system
unavailability?

l= 1.5 f/yr r = 10 hr = 10/8760 yr

 = 1/r = 8760/10 = 876 repairs/yr

Unavailability

U = l/(l+) = 1.5/(1.5+876) = 0.00171

= 0.00171 x 8760 = 14.97 hr/yr

58
 Availability Example – Series System
A generator supplies power through a transmission line. The failure rate
and the average repair time of the generator are 4 failures/year and 60
hours respectively, and that of the line are 2 failures/year and 10 hours
respectively. What is the unavailability of power supply?

G L

AG AL
Generator: Transmission Line:
lG= 4 f/yr lL= 2 f/yr
G= 1/rG = 8760/60 = 146 rep/yr L= 1/rL = 8760/10 = 876 rep/yr
AG = 1/(lG+G) = 146/(4+146) AL = L/(lL+L) = 876/(2+876)
= 0.973333 = 0.997722
Availability of the series system, Asys = AG x AL
= 0.973333 x 0.997722 = 0.971116
System Unavailability, Usys = 1 – Asys = 1 – 0.971116 = 0.028884
= 0.028884 x 8760 = 253.0 hr/yr
59
 Frequency and Duration Evaluation

Frequency of encountering State i

1
= P(being in State i) x (rate of departure from State i)
Up
= P(not being in State i) x (rate of entry into State i)
l 
2 P1.l = P2. .. Eq. 1
Down P1 + P2 = 1 .. Eq. 2

Solving Equations 1 and 2, P1 = = A and P2 = =U

Frequency of encountering the Down State,

FDown = P2 x (rate of departure from State 2) =

Mean Duration in the Down State = U / FDown = 1/

60
 Frequency and Duration Evaluation
lB
1 1 A (Up) 3 A (Up)
Up B (Up) B (Dn)
B
l  lA A lA A
2 lB
2 A (Dn) 4 A (Dn)
Down
B (Up) B (Dn)
B

Probability of being in State i  Availability, Unavailability

Frequency of encountering State i

= P(being in State i) x (rate of departure from State i)

Probabilit y of being in State i

Mean Duration in State i =
Frequency of encounteri ng State i
61
 Parallel System Evaluation
μA μB
lB P1 = ( )( )
1 A 3 A λ A  μ A λ B  μB
B B B
λA μB
P2 = ( )( )
lA A lA A λ A  μ A λ B  μB
μA λB
lB P3 = ( )( )
2 A 4 A λ A  μ A λ B  μB
B B
B λA λB
P4 = ( )( )
λ A  μ A λ B  μB

λA λB
System Unavailability, U = P4 = ( )( )
λ A  μ A λ B  μB

Frequency of Failure
= (P4).(rate of departure from State 4) = U.(A + B)

Mean Duration of Failure = U / Ffailure = 1/ (A + B)

62
 Parallel System Example
A customer is supplied by a distribution system that consists of an underground
cable in parallel with an overhead line. The failure rate and the average repair
time of the cable are 1 failure/year and 100 hours respectively, and that of the
overhead line are 2 failure/year and 10 hours respectively. Evaluate the
unavailability, frequency and the mean duration of failure of the distribution
system.

Underground Cable: Overhead Line:

lA= 1 f/yr lB= 2 f/yr
A= 1/r1 = 8760/100 = 87.6 rep/yr B= 1/r2 = 8760/10 = 876 rep/yr

λA λB
System Unavailability, U = P4 = ( )( )
λ A  μ A λ B  μB
= [1/(1+87.6)].[2/(2+876)] = 0.000026
= 0.000026 x 8760 = 0.2252 hr/yr

Frequency of Failure = U.(A + B) = 0.000026 x (87.6 + 876) = 0.0251 f/yr

Mean Duration of Failure = 1/ (A + B) = 1/(87.6 + 876) = 0.001 yr = 9.09 63

hr
 Series System Evaluation
lB Component A: lA= 1 f/yr, A= 87.6 r/yr
1 A 3 A Component B: lB= 2 f/yr, B= 876 r/yr
B B B

lA A lA A P1 = 0.986461
P2 = 0.011261
lB P3 = 0.002252
2 A 4 A
B B P4 = 0.000026
B

System Unavailability, U = P2 + P3 + P4 = 0.013539

= 0.013539 x 8760 = 118.60 hr/yr

Frequency of Failure, Ffailure = P2.A + P3.B

= 0.011261 x 87.6 + 0.002252 x 876 = 2.96 f/yr

Mean Duration of Failure = U / Ffailure = 0.013539 / 2.96 = 0.004575 yr

= 0.004575 x 8760 = 40.08 hr
64
 Modeling Failure, Repair, Installation

1
1
Up
Up g System Up

l  l System Down
2 2 3 Repaired but
Down Failed  not installed

P1.l = P3.g
P2. = P1.l
P1 + P2+ P3 = 1

Unavailability, U = P2+ P3
Frequency of encountering the Down State, FDown = P3 .g
Mean Duration in the Down State = U / FDown

65
 Modeling Spares and Installation Process

1 μγ
P1 =
Up λμ  λγ  μγ
g
λγ
l P2 =
λμ  λγ  μγ
2 3 Repaired
Failed but not λμ
 installed
P3 =
λμ  λγ  μγ

System Unavailability, U = P2 + P3

λμγ
Frequency of Failure, Ffailure = P3.g 
λμ  λγ  μγ

Mean Duration of Failure = U / Ffailure = (1/ ) + (1/ g)

66
 Modeling Failure, Repair, Installation

Example: A 138 kV, 40 MVA transformer has a failure rate of 0.1625 f/yr, and
average repair and installation times of 171.4 hours and 48 hours respectively.

μγ
P1 = λμ  λγ  μγ = 0.995946
l= 0.1625 f/yr
= 1/r = 8760/171.4 = 51.1 r/yr λγ
g = 8760/48 = 182.5 P2 = = 0.003167
λμ  λγ  μγ
λμ
P3 = = 0.000887
λμ  λγ  μγ

Unavailability, U = P2+ P3 = 0.004053 = 35.50 h/yr

Frequency of encountering the Down State, FDown = P3 .g = 0.1619 f/yr
Mean Duration in the Down State = U / FDown = 219.3 h

67
 Spare Component Assessment

4  1

0 spare 1 spare System Up

g g

l l System Down
5 2 3

0 spare 2 1 spare  2 spares

Unavailability, U = P2+ P3 + P5 = 1 – (P1+ P4)

Frequency of encountering the Down State, FDown = (P2 + P3).g = (P1 + P4).l
Mean Duration in the Down State = U / FDown

68
 Spare Assessment Example
Example: A 138 kV, 40 MVA transformer has a failure rate of 0.1625 f/yr, and
average repair and installation times of 171.4 hours and 48 hours respectively.
An identical spare is available.

l= 0.1625 f/yr P1 = 0.9966326

= 1/r = 8760/171.4 = 51.1 r/yr
g = 8760/48 = 182.5 P4 = 0.0024738

Unavailability, U = 1 – (P1+ P4) = 0.0008936 = 7.828 h/yr

Frequency of encountering the Down State, FDown = (P1 + P4).l = 0.1623
f/yr
Mean Duration in the Down State = U / FDown = 48.23 h

69
 F & D Using Approximate Equations

U = Ffailure .r

 l.r for MTTF (1/l)  MTBF (1/ Ffailure)

Up MTTF
MTBF MTTR
Down
0 Time

70
 Practical Adequacy Indices

• Failure rate (or frequency)

λ= failures/operating time
f = failures/time
• Average outage time
r = time/failure
• Average annual outage time
U = f.r ≈λ.r

71
 Series Systems

1 2 S

ls  l1  l 2   li
l1r1  l2 r2  l1l2 r1r2
rs 
l1  l2
l1r1  l2 r2  li ri
 
l1  l2  li
U s  ls rs
72
 Parallel Systems
1
S
2

l1l2 (r1  r2 )
ls 
1  l1r1  l2 r2
 l1l2 (r1  r2 )
r1r2
rs 
r1  r2
U s  ls rs 73
 Availability, F & D – Series System

1 2

Component 1: Component 2:
l1 = 1 f/yr l2= 2 f/yr
r1 = 100 hr r2 = 10 hr

System failure rate, ls =  li = l1 + l2 = 1 + 2 = 3 f/yr

System unavailability, Us =  li ri = 1 x 100 + 2 x 10 = 120 hr/yr

System average down time, rs = Us/ ls = 120/3 = 40 hr

74
 Availability, F & D – Parallel System
1

2
Component 1: Component 2:
l1 = 1 f/yr l2= 2 f/yr
r1 = 100 hr r2 = 10 hr

System failure rate, ls = l1.l2 (r1 + r2)

= 1 x 2 x (100 + 10)/8760 = 0.0251 f/yr
System average down time, rs = r1.r2 / (r1 + r2)
= 100 x 10 / (100 + 10) = 9.09 hr
System unavailability, Us = ls rs = 0.025 x 9.09 = 0.228 hr/yr
75
 Approximate Equations for Parallel Systems
l1
1
1

2
l2
2

For a 2-component parallel system,

ls  l1.l2 (r1 + r2) for li .ri << 1  ls = l1(l2 r1) + l2(l1 r2)
rs = r1.r2 / (r1 + r2)  s =  i
Us = ls .rs

76
 Similar equations can be used to
incorporate:
• Forced outages overlapping maintenance
outages
• Temporary outages
• Common mode outages
• Failure bunching due to adverse weather.

77
Forced outages overlapping maintenance
outages
λpm = λ1′′(λ2 r1′′) + λ2′′ (λ1r2′′)
Upm = λ1′′ (λ2r1′′) (r1′′r2)/(r1′′+ r2)
+ λ2′′ (λ1r2′′) (r1r2′′)/(r1+ r2′′)
rpm = Upm / λpm
where: λ′′ = maintenance outage rate
r′′ = maintenance time

78
 Basic Network Analysis Techniques

• Series / Parallel
2
Reduction
1
3
• Minimal Cut Set
Analysis

79
 Minimal Cut Set Analysis

2 l1  0.1 f / yr r1  100hrs
1
3
l2  l3  3 f / yr r2  r3  8hrs
Min Cuts l r U
1 0.1 100 10.0000
2,3 0.0164 4 0.0656
Total 0.1164 86.47 10.0656
ls  0.1164 f / yr
rs  86.47hrs
U s  10.0656hrs / yr
80
 Monte Carlo Simulation
Reliability Evaluation Techniques:
Analytical Technique Simulation Technique

represent the system by a simulate the actual process (using

mathematical model (usually random numbers) over the period of
simplified for practical systems) interest
direct mathematical solution repeat simulation for a large number of
times until convergence criteria is met
Advantages:

short solution time can incorporate complex systems

(analytical approach simplification can be
same results for the same unrealistic)
problem (greater but perhaps
unrealistic confidence to user) wide range of output parameters
including probability distributions
(analytical approach usually limited to
expected values) 81
 MCS Methods

Random Simulation
Basic (time) intervals chosen randomly
Can be applied when events in one basic interval do not affect the other
basic intervals

Sequential Simulation
Basic (time) intervals in chronological order
Required when one basic interval has a significant effect on the next
interval
Can also provide frequency and duration indices

82
 Random Simulation
1

U
1 2

A1= 0.8 A2= 0.6

A

U = Random # (0 – 1) 0
Simulation Convergence

Trial Component 1 Component 2 System System

# simulation simulation State Availability
Rand # State Rand # State
1 0.12 Up 0.35 Up Up 1/1 = 1.00
2 0.87 Down 0.21 Up Down 1/2 = 0.50
3 0.95 Down 0.62 Down Down 1/3 = 0.33
4 0.59 Up 0.18 Up Up 2/4 = 0.50
5 83
 Inverse Transform Method

An exponential variate T has the density function:

fT (t) = le-lt
Using the inverse transform method:
U is a uniform random number in the range of (0, 1).
U = FT (T) = 1 – e-lt
T  1 ln (1 – U)
l
= 1 ln U
l

84
 Sequential Simulation

1 Component 1: Component 2:
l1 = 1 f/yr l2= 5 f/yr

2 Evaluate the system reliability for an operating

time of 20 hours.
1
Up time = - ln U
λ

5
4
3
# of Simulations
2
1

0 Time (h) 20
85
 Sequential Simulation

1 Component 1: Component 2:
l1 = 1 f/yr l2= 5 f/yr
r1 = 100 hr r2 = 444 hr
2
Usys = U1 x U2 = 0.00228 = 20 hr/yr

1
Up time = - ln X Up >>
λ
1
Down time = - ln X
 Down >>
0 Time

total outage time total # of failures

U= Frequency of Failure = total simulation time
total simulation time

total outage time

Duration of Failure =
total # of failures
86
 Monte Carlo / Analytical Methods

• Monte Carlo simulation is a very

powerful approach and can be used to
solve a wide range of problems.
• In many cases, a suitable solution can be
obtained by using a direct analytical
technique.
• Use the most appropriate method for the
given problem

87
Basic Probability and Reliability Concepts

1. Develop an expression for the reliability of

the following system. Calculate the system
reliability if all the components have a
reliability of 0.8. 3/5 system
5
Redundant

2 3 5
1
5
4

5
6
5
Redundant 0
 Basic Probability and Reliability Concepts

RS  [R1 (R2 R3  R4  R2 R3 R4 )  R6  R1R6 (R2 R3  R4  R2 R3 R4 )]

5 4 3 2
 [R5  5R5 Q5  10R5 Q5 ]

R  0.8

RS  [0.8(0.928)  0.8  0.64(0.928)][0.942080]

 [0.948480][0.94208]  0.893544

1
Basic Probability and Reliability Concepts

2. (a) Calculate the availability of the following

system if each component has a failure rate of
5 f/yr and an average repair time of 92.21
hours.
1 2 3

Input 4 Output

5 6

(b) Estimate the system availability using

minimal cut sets. 2
Basic Probability and Reliability Concepts
Rs=Rs(4 is good)R4 + Rs(4 is bad)Q4
Given 4 is good
Rs=Rs(3 is good)R3 + Rs(3 is bad)Q3
=(R1 + R5 - R1R5) R3 + (R5R6) Q3
Given 4 is bad
Rs=R1R2R3 + R5R6 - R1R2R3R5R6

Substituting
Rs=R4[(R1 + R5 - R1R5) R3 + (R5R6) Q3]
+Q4[R1R2R3 + R5R6 - R1R2R3R5R6]
3
Basic Probability and Reliability Concepts

 5
Component Unavailability = Q =   0.05
   5  95

System availability = (0.95)[0.99275] + (0.05)[0.986094]


= 0.992417

System Unavailability = 0.007583

4
Basic Probability and Reliability Concepts

Min Cuts Probability

1, 5 0.0025
3, 5 0.0025
3, 6 0.0025
2, 4, 5 0.000125
2, 4, 6 0.000125
1, 4, 6 0.000125
System Unavailability ≤0.007875
System Availability ≥0.992125
5
Basic Probability and Reliability Concepts

1. Develop an expression for the reliability of

the following system. Calculate the system
reliability if all the components have a
reliability of 0.8. 3/5 system
5
Redundant

2 3 5
1
5
4

5
6
5
Redundant 0
Basic Probability and Reliability Concepts

2. (a) Calculate the availability of the following

system if each component has a failure rate of
5 f/yr and an average repair time of 92.21
hours.
1 2 3

Input 4 Output

5 6

(b) Estimate the system availability using

minimal cut sets. 1
Generating Capacity Reliability Evaluation

1. A generating system contains three 25 MW

generating units each with a 4% FOR and one 30
MW unit with a 5% FOR. If the peak load for a
100 day period is 75 MW, what is the LOLE and
LOEE for this period. Assume that the appropriate
load characteristic is a straight line from the 100%
to the 60% point.

0
Generating Capacity Reliability Evaluation
3 - 25 MW units 1 - 30 MW units
U = 0.04 U = 0.05
Cap Out Probability Cap Out Probability
0 0.884736 0 0.95
25 0.110592 30 0.05
50 0.004608 1.000000
75 0.000064
1.000000

1
Generating Capacity Reliability Evaluation

IC=105 MW

75 MW

45 MW

0 100 days

2
Generating Capacity Reliability Evaluation
Total Capacity
Cap Out Probability Time (hrs) Energy (MWh)
0 0.840499 --
25 0.105062 --
30 0.044237 --
50 0.004378 1600 16,000
55 0.005530 2000 25,000
75 0.000061 2400 72,000
80 0.000230 2400 84,000
105 0.000003 2400 144,000
1.0
3
Generating Capacity Reliability Evaluation

n
LOLE   pk tk = 18.77 hrs/100d period
k 1

n
LOEE   p k E k = 232.44 MWh / 100 day period
k 1

4
Generating Capacity Reliability Evaluation

• Loss of Load Expectation, LOLE = 18.77 hrs/100 d

period
• Loss of Energy Expectation, LOEE = 232.44 MWh/100
d period
232.44
• Energy Index Reliability EIR = 1  0.998386
144,000
• Energy Index of Unavailability EIU = 0.001614
• Units per Million UPM= 1614
232.44
• System Minutes SM =  60  185.95
75
5
Generating Capacity Reliability Evaluation

2. Two power systems are interconnected by a 20 MW

tie line. System A has three 20 MW generating units
with forced outage rate of 10%. System B has two 30
MW units with forced outage rates of 20%. Calculate
the LOLE in System A for a one-day period, given
that the peak load in both System A and System B is
30 MW.
20 MW
A B

3-20 MW 2-30 MW
U=0.1 U=0.2
L=30 MW L=30 MW 6
Generating Capacity Reliability Evaluation

System A System B
Cap Out Probability Cap Out Probability
0 0.729 0 0.64
20 0.243 30 0.32
40 0.027 60 0.04
60 0.001 1.00

7
Generating Capacity Reliability Evaluation

Capacity Array Approach

System B
0 30 60
System A 0 0.46656 0.23328 0.02916
20 0.15552 0.07776 0.00972
40 0.01728 0.00864 0.00108
60 0.00064 0.00032 0.00004

LOLE(A)[Single System] = 0.028 days/day

LOLE(A)[Interconnected System] = 0.01072 days/day
8
Generating Capacity Reliability Evaluation
Equivalent Unit Approach
20 MW Assisting Unit Modified System A IC = 80 MW

Cap Out Probability Cap Out Probability Cum. Probability

0 0.64 0 0.46656 1
20 0.36 20 0.41796 0.53344
40 0.10476 0.11548
60 0.01036 0.01072
80 0.00036 0.00036
1.000000

LOLE(A)[Interconnected System] = 0.01072 days/day

9
Generating Capacity Reliability Evaluation

1. A generating system contains three 25 MW

generating units each with a 4% FOR and one 30
MW unit with a 5% FOR. If the peak load for a
100 day period is 75 MW, what is the LOLE and
LOEE for this period. Assume that the appropriate
load characteristic is a straight line from the 100%
to the 60% point.

0
Generating Capacity Reliability Evaluation

2. Two power systems are interconnected by a 20 MW

tie line. System A has three 20 MW generating units
with forced outage rate of 10%. System B has two 30
MW units with forced outage rates of 20%. Calculate
the LOLE in System A for a one-day period, given
that the peak load in both System A and System B is
30 MW.
20 MW
A B

3-20 MW 2-30 MW
U=0.1 U=0.2
L=30 MW L=30 MW 1
 Transmission System Reliability Evaluation

1. Consider the following system

1 2 B
4
Supply A C
3

The supply is assumed to have a failure rate of 0.5 f/yr

with an average repair time of 2 hours. The line data are
as follows.

1
Transmission System Reliability Evaluation

Line Failure Rate Average Repair

Time
1 4.0 f/yr 8 hrs
2 2.0 6
3 6.0 8
4 2.0 12
Use the minimal cut set approach to calculate a
suitable set of indices at each load point.
2
 Transmission System Reliability Evaluation

Load Point A
Min Cut  (f/yr) r (hrs) U (hrs/yr)
Supply 0.5 2.0 1.0
1, 3 0.043836 4.0 0.175344

1, 2 0.012785 3.4286 0.043835
0.556621 2.19 1.219179
1 2 B
4
Supply A C
3

3
 Transmission System Reliability Evaluation

Load Point B
Min Cut  (f/yr) r (hrs) U(hrs/yr)
Supply 0.5 2.0 1.0
1, 3 0.043836 4.0 0.175344

2, 3 0.019178 3.4285 0.065753
0.563014 2.2044 1.241097
1 2 B
4
Supply A C
3

4
 Transmission System Reliability Evaluation

Load Point C
Min Cut  (f/yr) r (hrs) U(hrs/yr)
At B 0.563014 2.2044 1.241097
4 2.0 12 24

2.563014 9.848 25.241097

1 2 B
4
Supply A C
3

5
 Transmission System Reliability Evaluation

Summary
Min Cut  (f/yr) r (hrs) U (hrs/yr)
A 0.5566 2.19 1.219
B 0.5630 2.20 1.241

C 2.5630 9.85 25.241

1 2 B
4
Supply A C
3

6
Composite System Reliability
Evaluation
2. A four unit hydro plant serves a remote load through
two transmission lines. The four units are connected to a
single step-up transformer which is then connected to two
transmission lines. The remote load has a daily peak load
variation curve which is a straight line from the 100% to
the 60% point. Calculate the annual loss of load
expectation for a forecast peak of 70 MW using the
following data.
Hydro Units – 25 MW Transformer – 110 MVA
FOR = 2% U = 0.2%
Transmission lines – Carrying capability 50 MW per line
– Failure rate = 2 f/yr
– Average repair time = 24 hrs 7
 Composite System Reliability
Evaluation
Calculate the LOLE in three stages using the
following configurations.

(a) (c)

(b)

8
(d) Calculate the LOLE for Configuration (b), if the single
step-up transformer is removed and replaced by
individual unit step-up transformers with a FOR of
0.2%.
(e) Calculate the LOLE for the conditions in (d) with each
transmission line rated at 50 MW.
(f) Calculate the LOLE for the conditions in (d) with each
transmission line rated at 75 MW.
(g) Calculate the LOLE for the conditions in (d) with each
transmission line rated at 100 MW.
(h) Calculate the LOLE for the conditions in (f) with Model
1 common mode TL failure. [ λc = 0.2 f/yr ]
(i) Calculate the LOLE for the conditions in (f) with Model
3 common mode TL failure. [ λc = 0.2 f/yr, rc = 36 hr ]

9
Composite System Reliability Evaluation

Configuration (a)
Capacity Out Probability Time Expectation
0 MW 0.922368 0.0
25 0.075295 0.0
50 0.002305 260.71 0.600937
75 0.000032 365.0 0.011680
100 - 365.0 -
1.000000 0.612617

LOLE = 0.613 days/yr 10

Composite System Reliability Evaluation

Configuration (b)
Capacity Out Probability Time Expectation
0 MW 0.920524 0.0
25 0.075144 0.0
50 0.002300 260.71 0.599633
75 0.000032 365.0 0.011680
100 0.002000 365.0 0.730000
1.000000 1.341313

LOLE = 1.341 days/yr 11

Composite System Reliability Evaluation

Configuration (c)
Transmission lines 2 f / yr
1 8760
   365 r / yr
r 24
  2
Unavailability    0.005450
   2  365
 Cap. Out Probability
Availability  0.994550
0 MW 0.989130
50 0.010840
100 0.000030
1.000000 12
Composite System Reliability Evaluation

Generation – In (MW)
T/G 100 75 50 25 0

100 100 75 50 25 0

Transmission-In 50 50 50 50 25 0
(MW)
0 0 0 0 0 0

System Capacity States

13
Composite System Reliability Evaluation
Configuration (c)

Capacity Probability Time Expectation

In Out
100 0 0.910518 0.0
75 25 0.074327 0.0
50 50 0.013093 260.71 3.413476
25 75 0.000032 365.0 0.011680
0 100 0.002030 365.0 0.740950
1.000000 4.166106

LOLE = 4.166 days/yr 14

Composite System Reliability Evaluation

Configuration (d)
Calculate the LOLE for Configuration (b), if the
single step-up transformer is removed and replaced
by individual unit step-up transformers with a FOR
of 0.2%.

Generating unit FOR = 0.02 + 0.002 – (0.02)(0.002)

U = 0.021960
A = 0.978040
15
Composite System Reliability Evaluation
Configuration (d)

Capacity Probability Time Expectation

In Out
100 0 0.915012 0.0
75 25 0.082179 0.0
50 50 0.002768 260.71 0.721645
25 75 0.000041 365.0 0.014965
0 100 - 365.0 -
1.000000 0,733661

LOLE = 0.734 days/yr 16

(e) Calculate the LOLE for the conditions in (d) with each
transmission line rated at 50 MW.

Capacity Probability Time Expectation

In Out
100 0 0.905066 0.0
75 25 0.081286 0.0
50 50 0.013577 260.71 3.539660
25 75 0.000041 365.0 0.014965
0 100 0.000030 365.0 0.010950
1.000000 3.565575

LOLE = 3.566 days/yr 17

(f) Calculate the LOLE for the conditions in (d) with each
transmission line rated at 75 MW.

Capacity Probability Time Expectation

In Out
100 0 0.905066 0.0
75 25 0.092095 0.0
50 50 0.002768 260.71 0.721645
25 75 0.000041 365.0 0.014965
0 100 0.000030 365.0 0.010950
1.000000 0.747550

LOLE = 0.748 days/yr 18

(g) Calculate the LOLE for the conditions in (d)
with each transmission line rated at 100 MW.

Capacity Probability Time Expectation

In Out
100 0 0.914985 0.0
75 25 0.082177 0.0
50 50 0.002768 260.71 0.721645
25 75 0.000041 365.0 0.014965
0 100 0.000030 365.0 0.010950
1.000000 0.747550

LOLE = 0.748 days/yr 19

(h) Calculate the LOLE for the conditions in (f) with Model 1
common mode TL failure.

λc
2λ λ
Both One Up Both
UP One Down Down
µ 2µ

P(Both Up) = 0.988326

P(One Up and One Down = 0.011372
P(Both Down) = 0.000302
20
 21

Markov analysis of Model 1

P4 = [λ1 λ2 (λ1 + λ2 + 1 + 2) +
λc (λ1 + 2)(λ2 + 1)] / D
D = (λ1 + 1)(λ2 + 2)(λ1 + λ2 + 1 + 2)
+ λc[(λ1 + 1)(λ2 + 1 + 2) + 2 (λ2 + 2)]

If the two components are identical

P4 = [2λ2 + λc (λ + )] / [2(λ + )2 + λc (λ+ 3)]

= P(Both Down) = 0.000302

21
The basic reliability indices for Model 1 can be
estimated using an approximate method [1].

System failure rate = λs = λ1 λ2 (r1 + r2) + λc

Average system outage time = rs = (r1 r2) /(r1 + r2)
System unavailability = Us = λs rs

P( Both Down) = 0.000304

22
Approximate calculation for:

P(One line Up & One line Down) = 2 AL. UL

= 2.(2/367)(365/367)
= 0.010840

P(Both lines Up) = 1.0 - 0.010840 – 0.000304

= 0.988856

Combine the generation and transmission states.

LOLE = 0.847310 days/year

23
Approximate method applied to Model 3
In this case:
λs = λ 1 λ2 ( r 1 + r 2 ) + λ c
Us = λ1 λ2 r1 r2 + λc rc
rs = Us / λs

P( Both Down) = 0.000852

24
Approximate calculation for:

P(One line Up & One line Down) = 2 AL. UL

= 2.(2/367)(365/367)
= 0.010840

P(Both lines Up) = 1.0 - 0.010840 – 0.000852

= 0.988308

Combine the generation and transmission states.

LOLE = 1.47069 days/year

25
Composite System Reliability Evaluation
Conditions LOLE d/y
(a) Generation (G) only 0.613
(b) (G) with single transformer (T) 1.341
(c) G, T and two 50 MW transmission lines 4.166
(d) (G) with unit transformers 0.734
(e) Generation only 0.613
(f) Condition (d) with two 50 MW transmission lines 3.566
(g) Condition (d) with two 75 MW transmission lines 0.748
(h) Condition (d) with two 100 MW transmission lines 0.748
(i) Condition (f) with Model 1 common mode TL failure 0.847
(j) Condition (f) with Model 3 common mode TL failure 1.471

26
Composite System Reliability
Evaluation
2. Consider the following system
1 2
1

A
2 3

1. Calculate the probability of load curtailment

at load points A and B
2. Calculate the EENS at load points A and B
27
27
Composite System Reliability Evaluation
• System Data
Generating Stations
1. 4*25 MW units   2.0 f / yr   98.0r / yr
2. 2*40 MW units   3.0 f / yr   57.0r / yr
Loads

A 80 MW

B 60 MW
Transmission Lines
1   4 f / yr, r  8hrs, LCC  80MW
2   5 f / yr, r  8hrs, LCC  60MW
3   3 f / yr, r  12hrs, LCC  50MW 28
28
Composite System Reliability Evaluation

• Conditions
– Assume that the loads are constant
– Assume that the transmission loss is zero
– Consider up to two simultaneous outages
– Assume that all load deficiencies are
shared equally where possible.

29
Composite System Reliability Evaluation

• Element Probabilities
Element   /r A U
25 MW unit 2.0 f/yr 98.0 r/yr 0.98 0.02
40 MW unit 3.0 57.0 0.95 0.05
 
L1 4.0 8 hrs 0.99636033 0.00363967
L2 5.0 8 0.99545455 0.00454545
L3 3.0 12 0.99590723 0.00409277

30
Composite System Reliability
Evaluation
• Plant Probabilities
Conditions P(Plant 1) P(Plant 2)
All Units In 0.92236816 0.90250
1 Unit Out 0.07529536 0.09500
2 Unit Out 0.00230496 0.00250

All Lines In 0.98777209

31
Basic Structure: Base case analysis
Select a contingency
Simulation Evaluate the selected contingency
Sample No
There is a system problem
Load Yes
Generators Take appropriate remedial action
Weather
Transmission No
There is still a system problem
Trials Yes
complete? Evaluate the impact of the problem
Calculate and summate the load
point reliability indices
Compile overall Yes No
All contingencies evaluated
system indices 32
Composite System Reliability Evaluation

4*25 2*40
(100 MW) 1 2 (80 MW)
1 (80 MW)

A (80 MW)
2 3
(60 MW) (50 MW)

Total Cap. 180 MW B (60 MW)

Total Load 140 MW

33
 Composite System Reliability Evaluation
State Condition A B State Condition A B
1 No Outages -- -- 10 1 G2, L1 × ×
2 1 G1 -- -- 11 1 G2, L2 × ×
3 1 G1, 1 G1 × × 12 1 G2, L3 -- --
4 1 G1, 1 G2 × × 13 L1, -- --
5 1 G1, L1 -- -- 14 L1, L2 × ×
6 1 G1, L2 -- × 15 L1, L3 -- --
7 1 G1, L3 -- -- 16 L2 -- ×
8 1 G2, -- -- 17 L2, L3 -- ×
9 1 G2, 1 G2 × × 18 L3, -- --

34
Composite System Reliability Evaluation
State Condition Probability LC EENS
3 G1, G1 0.002055 5 MW 90.01 MWh/yr
4 G1, G2 0.007066 12.5 773.73
9 G2, G2 0.002278 20 399.11
10 G2, L1 0.000316 20 55.36
11 G2, L2 0.000395 10 34.60
14 L1, L2 0.000014 30 3.68
0.012124 1356.49
U(A) = 0.012124
EENS(A) = 1356.49 MWh/yr 35
Composite System Reliability Evaluation
State Condition Probability LC EENS
3 G1, G1 0.002055 5 MW 90.01 MWh/yr
4 G1, G2 0.007066 12.5 773.73
6 G1, L2 0.000307 10 26.89
9 G2, G2 0.002278 20 399.11
10 G2, L1 0.000316 20 55.36
11 G2, L2 0.000395 10 34.60
14 L1, L2 0.000014 30 3.68
16 L2 0.003755 10 328.94
17 L2, L3 0.000015 60 7.88
0.016201 1720.20
U(B) = 0.016201
36
EENS(B) = 1720.20 MWh/yr
 Transmission System Reliability Evaluation

1. Consider the following system

1 2 B
4
Supply A C
3

The supply is assumed to have a failure rate of 0.5 f/yr

with an average repair time of 2 hours. The line data are
as follows.

1
Transmission System Reliability Evaluation

Line Failure Rate Average Repair

Time
1 4.0 f/yr 8 hrs
2 2.0 6
3 6.0 8
4 2.0 12
Use the minimal cut set approach to calculate a
suitable set of indices at each load point.
2
Composite System Reliability
Evaluation
2. A four unit hydro plant serves a remote load through
two transmission lines. The four units are connected to a
single step-up transformer which is then connected to two
transmission lines. The remote load has a daily peak load
variation curve which is a straight line from the 100% to
the 60% point. Calculate the annual loss of load
expectation for a forecast peak of 70 MW using the
following data.
Hydro Units – 25 MW Transformer – 110 MVA
FOR = 2% U = 0.2%
Transmission lines – Carrying capability 50 MW per line
– Failure rate = 2 f/yr
– Average repair time = 24 hrs 3
 Composite System Reliability
Evaluation
Calculate the LOLE in three stages using the
following configurations.

(a) (c)

(b)

4
(d) Calculate the LOLE for Configuration (b), if the single
step-up transformer is removed and replaced by
individual unit step-up transformers with a FOR of
0.2%.
(e) Calculate the LOLE for the conditions in (d) with each
transmission line rated at 50 MW.
(f) Calculate the LOLE for the conditions in (d) with each
transmission line rated at 75 MW.
(g) Calculate the LOLE for the conditions in (d) with each
transmission line rated at 100 MW.
(h) Calculate the LOLE for the conditions in (f) with Model
1 common mode TL failure. [ λc = 0.2 f/yr ]
(i) Calculate the LOLE for the conditions in (f) with Model
3 common mode TL failure. [ λc = 0.2 f/yr, rc = 36 hr ]

5
Composite System Reliability
Evaluation
2. Consider the following system
1 2
1

A
2 3

1. Calculate the probability of load curtailment

at load points A and B
2. Calculate the EENS at load points A and B
6
6
Composite System Reliability Evaluation
• System Data
Generating Stations
1. 4*25 MW units   2.0 f / yr   98.0r / yr
2. 2*40 MW units   3.0 f / yr   57.0r / yr
Loads

A 80 MW

B 60 MW
Transmission Lines
1   4 f / yr, r  8hrs, LCC  80MW
2   5 f / yr, r  8hrs, LCC  60MW
3   3 f / yr, r  12hrs, LCC  50MW 7
7
Composite System Reliability Evaluation

• Conditions
– Assume that the loads are constant
– Assume that the transmission loss is zero
– Consider up to two simultaneous outages
– Assume that all load deficiencies are
shared equally where possible.

8
Probability Fundamentals and Models in
Generation and
Bulk System Reliability Evaluation

Roy Billinton
Power System Research Group
University of Saskatchewan
CANADA

1
Mission Reliability
Reliability is the probability of a
device or system performing its
purpose adequately for the period of
time intended under the operating
conditions encountered.
C.R. Knight, E.R. Jervis, G.R. Herd, “Terms of Interest in the
Study of Reliability”, IRE Transactions on Reliability and Quality
Control. Vol. PGRQC-5, April 1955, pp. 34-56.

2
Reliability
A measure of the ability of the system
to perform its intended function

Reliability Assessment
Deterministic
Probabilistic

3
 Deterministic - adjective

To determine:
 to fix  % Reserve
 to resolve ( N-1 )
 to settle Worst case
 to regulate condition
 to limit
 to define

4
 Probabilistic - adjective

Probability – likelihood of an event, the

expected relative frequency of
occurrence of a specified event
in a very large collection of
possible outcomes.

5
Probability  a quantitative measure of the
likelihood of an event.

 a quantitative measure of the

uncertainty associated with the
event occurring.
 a quantitative indicator of
uncertainty.
6
Probability concepts provide the ability to
quantitatively incorporate uncertainty in power
system planning applications.

This cannot be done using deterministic

methods and criteria.

7
Power system reliability assessment is usually
divided into the two areas of Adequacy and
Security evaluation
• Adequacy is generally considered to be the
existence of sufficient facilities within the
system to satisfy the consumer demand.

• Security is considered to relate to the ability

of the system to respond to disturbances
arising within that system.

8
 Incremental Reliability
1.0
System Reliability

∆R
∆C

System Cost
What is the system reliability benefit for the next dollar invested?

This requires a quantitative evaluation of system reliability.

9
Value Based Reliability Assessment
(VBRA) is a useful extension to
conventional reliability evaluation
and provides valuable input to the
decision making process.

10
Reliability Cost/Worth

11
Ontario Energy Board stated that Ontario Hydro had
too high a level of generation system reliability.

Ontario Hydro conducted a series of studies in 1976

– 1979 to determine the customer costs associated
with electric power supply failures and produced:

“The SEPR Study: System Expansion Program

Reassessment Study” Final Report 1979

12
Functional Zones and Hierarchical
Levels

Generation
Facilities Hierarchical Level I
HL-I

Transmission Hierarchical Level II

Facilities HL-II

Distribution Hierarchical Level III

Facilities HL-III

13
• “Reliability Evaluation of Engineering
Systems, Second Edition”, R. Billinton
and R.N. Allan, Plenum Press, 1992.,
pp. 453.

88
 Generating Capacity
Reliability Evaluation

Roy Billinton
Power System Research Group
University of Saskatchewan
CANADA

1
Functional Zones and Hierarchical Levels

Generation
Facilities Hierarchical Level I
HL-I

Transmission Hierarchical Level II

Facilities HL-II

Distribution Hierarchical Level III

Facilities HL-III

2
 Hierarchical Level I – HL-I

Classical generating capacity planning

Task – plan a generating system to meet the

system load requirement as economically as
possible with an acceptable level of reliability.

3
 System Model

Generation Alternatives
• Fossil
Remote Capacity G • Hydro
• Nuclear
• Gas
GR • Renewables
External Capacity
Load Characteristics
Uncertainty

4
Conceptual Tasks in Reliability Evaluation
at HLI

Generation Load

Loss of Load Expectation (LOLE)

Risk Loss of Energy Expectation (LOEE)
Frequency & Duration (F&D)
Other Indices

5
Loss of Load Expectation (LOLE) is the expected
number of hours or days in a given period of time
that the load exceeds the available generation.

Loss of Energy Expectation (LOEE) is the

expected energy not supplied in a given period of
time due to the load exceeding the available
generation.

The LOLE and LOEE are long run average values

and are important indicators of HLI adequacy.

6
The basic component model used in most power
system reliability studies is the two state
representation shown in Fig. 2.


Up Down

Fig. 2. Two state component model

7
The model shown in Fig. 2 is a simple but
reasonably robust representation. The component
availability (A) and unavailability (U) (Forced
Outage Rate) are given by Equation (1).

μ
A
λμ
(1)

U
λ

 (Down Time)
λμ  (Up Time)   (Down Time)
8
 There are many variations and expansions of the model
shown in Fig. 2, particularly in research related studies and
developments. Some of these are:

 The inclusion of derated states in generating units.

 The four state model used to recognize the

conditional probability of failure associated with
peaking units.

 The three state model used to consider active and

passive failures of circuit breakers.

 The recognition of non-exponential state residence

time distributions and variable failure and repair
rates due to component aging, repair and
maintenance practices.

9
 Derated State Model

Full
Output

λ1 λ2
µ1 µ2

λ3
Partial
Failed
Output µ3

10
 Two-State Models

The unit derated state model can be reduced to a two-state

representation. The derated adjusted forced outage rate
(DAFOR) is used by the Canadian Electricity Association
(CEA) to represent the probability of a multi-state unit being
in the forced outage state. and is obtained by apportioning the
time spent in the derated states to the full up and down states.
This is known as the equivalent forced outage rate (EFOR) in
the NERC-GADS

11
The two state representation in which the unit is
available or unavailable for service is a valid
representation for base load units but does not
adequately represent intermittent operating units used to
meet peak load conditions. Peaking units are started
when they are needed and normally operate for relatively
short periods. The operation of peaking units can be
described by the frequency and duration of their service
and shutdown states and the transitions between these
states.

12
 Four-State Model

The IEEE Subcommittee on the Application of

Probability Methods proposed a four-state
model for peaking units.
This model includes reserve shutdown and
forced out but not needed states.

13
 Four-State Model
(1-Ps)/T
Reserve Shutdown In Service

State 0 State 1
1/D

1/r Ps/T 1/r 1/m

1/T
Forced Out but Not Forced Out When
Needed Needed
State 3 State 2
1/D

T=Average reserve shutdown time between periods of need.

D = Average in service time per occasion of demand.
Ps = Probability of starting failure.
m and r are the same as in the two-state model.
14
 The UFOP and The Demand Factor
• The Utilization Forced Outage Probability (UFOP) is
the probability of a generating unit not being available
when needed.
P2
UFOP 
P1  P2
• The demand factor f of a peaking unit is calculated as
follows.
P2 (1 / r  1 / T )
f  
P2  P3 1 / D  1 / r  1 / T
• Pi represents the probability of State i.

15
 The UFOP and The Demand Factor

The conventional forced outage rate is:

FOR = (FOH) / (SH + (FOH))

The conditional forced outage rate is:

UFOP = f (FOH) / (SH + f (FOH))

16
 Canadian Electricity Association
Equipment Reliability Information System
Components

• Generation Equipment Status Reporting

System
• Transmission Equipment Outage Reporting
System
• Distribution Equipment Outage Reporting
System

17
In Table 1:
FOR = Forced Outage Rate,
DAFOR = Derated Adjusted Forced Outage Rate; This is known as
EFOR in the NERC-GADS
DAUFOP = Derated Adjusted Utilization Forced Outage Probability;
This is known as EFORd in the NERC-GADS and is the conditional
probability of finding the unit in the modified down state given that
the system needs the unit.
Table 1
Generating Unit Unavailability Statistics
Unit Type FOR % DAFOR % DAUFOP %
Hydraulic 1.97 2.03 1.74
Fossil 7.32 10.74 9.16
Nuclear 7.64 9.16 9.12
CTU 29.78 ----- 8.13

where: CTU = Combustion Turbine Unit

18
 Table 2
FOR, DAFOR and DAUFOP for Hydraulic Units by Unit Size
MCR (MW) FOR (%) DAFOR (%) DAUFOP (%)
5 – 23 3.67 3.71 3.17
24 – 99 1.48 1.56 1.38
100 – 199 1.08 1.13 0.95
200 – 299 2.30 2.36 1.94
300 – 399 0.93 0.93 0.82
400 – 499 1.26 1.29 1.10
500 – over 0.64 0.64 0.59

Canadian Electricity Association “Generation Equipment

Status”, 2002-2006

19
 Table 3
FOR, DAFOR and DAUFOP for Fossil Units-Coal
by Years of Service
Years of Service FOR (%) DAFOR (%) DAUFOP (%)

6th – 10th 2.00 2.75 2.73

11th – 15th 2.06 2.89 3.25
16th – 20th 3.76 4.67 4.64
21st – 25th 4.26 6.22 6.10
26th – 30th 6.61 11.26 10.58
31st – 35th 9.26 13.57 12.82
36th – 40th 12.90 18.89 15.73
41st – 45th 12.69 17.15 13.99
46th – 50th 4.18 12.45 12.06

20
The unavailability statistics shown in Tables 1-3
are normally associated with adequacy
assessment and used in planning studies. The
most important parameters in an operating or
short-term sense is the generating unit failure
rate (λ). The probability of a unit failing in the
next few hours, Q(t), is given by Equation (4).
-λt
Q(t) = 1–e ≈ λ.t (4)

The assumption in this case is that the time

period t is sufficiently short that repair is not a
factor.
21
The λ.t term has been designated as the Outage
Replacement Rate (ORR) and is used as the basic
generating unit statistic in spinning or operating
reserve studies. Table 4 shows representative failure
rates for the general unit classes in Table 1.

Table 4
Generating Unit Failure Rates

Unit Type Failure Rate (f/a)

Hydraulic 2.30
Fossil 10.70
Nuclear 2.24
CTU 10.82

22
 Risk evaluation method and equations

Installed Capacity (MW) n

LOLE   pk tk
Reserve k 1
n
LOEE   p k E k
Area=Ek Outage k k 1

where
Time when there is loss of Load Curve n is the total number of capacity outage
load t k states.
p k is the individual probability of the
capacity outage state k.
t k is the number of time units when there
is a loss of load.
E k represents the energy that cannot be
supplied in a capacity outage state k.

0 Time 1

23
Monte Carlo Simulation
M

t k
LOLE  k 1
N
M

 ENS i
LOEE  i 1
MWh / yr
N

N: Sampling years
ENS i
M: Number of the
occurrence of Loss of Load
tk in N years.

24
 Generation Model
• Example: A 100 MW generating system consists of
five 20 MW units. Each unit has an FOR of 0.03.
• Binomial Distribution
Units Out Capacity Out Capacity In Individual Cumulative
(MW) (MW) Probability Probability
0 0 100 0.858734 1
1 20 80 0.132794 0.141266
2 40 60 0.008214 0.008472
3 60 40 0.000254 0.000258
4 80 20 0.000004 0.000004
5 100 0 0.000000 0.000000
25
 Load Model

• A load with a peak of 60 MW and a load factor

of 75%.

70
60
Load (MW)

50
40
30
20
10
0
0 8760
Time (hours)

26
 Risk Evaluation
100

Capacity and Load (MW)

X1=20 MW X2=40 MW X3=60 MW
80

60
X4=80 MW
40

20
X5=100 MW
0
0 5840 8760
Time (hours)

Cap. Out Cap. In Individual Outage LOL (hours/year) LOE (MWh/year)

(MW) (MW) Probability Time (hours) =C3*C4
0 100 0.858734 0 0 0
20 80 0.132794 0 0 0
40 60 0.008214 0 0 0
60 40 0.000254 5840 1.483360 14.8336
80 20 0.000004 8760 0.034427 0.8607
100 0 0.000000 8760 0.000213 0.0096
LOLE=1.5180 LOEE=15.7039
27
 LOLE versus Peak Load

• 5*20 MW Generating System

LOLE (Hours/year)

1000
100
10
1
0.1
0.01
40 45 50 55 60 65 70 75 80
Peak Load (MW)

FOR=0.03 FOR=0.05
28
 Add a 50 MW Unit, FOR=0.05
• Create a new COPT using the conditional probability method
2
P( A)   P( A | B j ) P( B j )
j 1 5*20 MW + 1*50 MW
5*20 MW FOR=0.03 Cap. Out Cap. In Individual
Cap. Out Cap. In Individual (MW) (MW) Probability
(MW) (MW) Probability 0 150 0.8157973
0 100 0.858734 20 130 0.1261542
20 80 0.132794 40 110 0.0078034
40 60 0.008214 50 100 0.0429367
60 40 0.000254 60 90 0.0002413
80 20 0.000004 70 80 0.0066397
100 0 0.000000 80 70 0.0000037

1*50 MW FOR=0.05 90 60 0.0004107

Cap. Out Cap. In Individual 100 50 0.0000000

(MW) (MW) Probability 110 40 0.0000127
0 50 0.95 130 20 0.0000002
50 0 0.05 150 0 0.0000000
29
 LOLE versus Peak Load

• 5*20 MW(FOR=0.03) Plus 1*50 MW(FOR=0.05)

LOLE (hours/year)

100
10
1
0.1
IPLCC=15MW
0.01
0.001
40 45 50 55 60 65 70 75 80
Peak Load (MW)

5*20 MW 5*20 MW Plus 1*50 MW

30
 Generation Models
Hydro and SCGT Units – 2 state models
Base load units – FOR, DAFOR
As needed unit - UFOP
CCGT Units – multi-state models for combined units

State Units Available

Probability
# Unavailable Capacity
1 none 2CGT + CST (1-FORST) x (1-FORGT)2
2 ST 2CGT FORST x (1-FORGT)2

3 1 GT CGT + 0.5CST 2 x (1-FORST) x (1-FORGT) x FORGT

4 1 GT + ST CGT 2 x FORST x (1-FORGT) x FORGT

5 2 GT 0 FORGT2

31
 Wind Power Modeling and Data

The power produced by a wind turbine

generator (WTG) at a particular site is highly
dependent on the wind regime at that location.
Appropriate wind speed data are therefore
essential elements in the creation of a suitable
WTG model. The actual data for a site or a
statistical representation created from the
actual data can be used in the model.
This is illustrated using data for a site located
at Swift Current in Saskatchewan, Canada.

32
The mean and standard deviation of the wind
speed at the Swift Current site are 19.46 km/h
and 9.7km/h respectively. The hourly mean and
standard deviation of wind speeds from a 20-
year database (1 Jan.1984 to 31 Dec. 2003) for
the Swift Current location were obtained from
Environment Canada. These data were used to
build an Auto-Regressive Moving Average
Model (ARMA) time series model.

33
The ARMA (4,3) model is the optimal time series model for
the Swift Current site and the parameters are shown in
Equation (1):
Swift Current: ARMA (4, 3):
yt  1.1772 yt 1  0.1001yt  2  0.3572 yt 3  0.0379 yt  4 (1)
 t  0.5030 t 1  0.2924 t  2  0.1317 t 3
 t  NID(0,0.5247602 )
The simulated wind speed SWt can be calculated from Equation (2)
using the wind speed time series model.
SWt  t   t  yt (2)
where µt is the mean observed wind speed at hour t,σt is the
standard deviation of the observed wind speed at hour t,  t  is a
normal white noise process with zero mean and the variance
0.5247602.

34
The hourly wind data produced by the ARMA
model can be used in a sequential Monte Carlo
simulation of the total system generation or to
create a multi-state model of the WTG that can
be used in an analytical technique or a non-
sequential Monte Carlo approach to generating
capacity assessment. A capacity outage
probability table (COPT) of a WTG unit can be
created by applying the hourly wind speed to
the power curve.

35
The power output characteristics of a WTG are
quite different from those of a conventional
generating unit and depend strongly on the
wind regime as well as on the performance
characteristics of the generator.
The parameters commonly used are the cut-in
wind speed (at which the WTG starts to
generate power), the rated wind speed (at
which the WTG generates its rated power) and
the cut-out wind speed (at which the WTG is
shut down for safety reasons) .

36
Wind Turbine Generating Unit Power Curve

37
Renewable energy sources, such as wind and solar power, behave quite
differently than conventional generation facilities.
Wind speeds & power outputs from two consecutively simulated years (the first week of January)

60
50 Year 1 Year 2
Wind Speed

40
(km/h)

30
20
10
0
1 25 49 73 97 121 145
Hour

Power Output (MW)

Pr
Power Curve Parameters:
Cut-in speed (Vci) = 14.4 km/h
Rated speed (Vr) = 36.0 km/h
Cut-out speed (Vco) = 80.0 km/h
Vci Vr Vco
Wind Speed (km/h)

50
Power Output

Year 1 Year 2
40
(MW)

30
20
10
0
1 25 49 73 97 121 145
Time (hour)

38
 0.05 1.2 0.05 1.2
Avg.= 19.4 km/h, S.D.= 10.1 km/h Avg.= 19.6 km/h, S.D.= 10.8 km/h

Power Output (p.u.)

Power Output (p.u.)

0.04 1 1
0.04
Probability

Probability
0.8 0.8
0.03 0.03
Probability 0.6 0.6
Probability
0.02 Power Output 0.4 0.02 Power Output 0.4
0.01 0.2 0.01 0.2
0 0 0 0
0 14 28 42 56 0 14 28 42 56
Year 1 Wind Speed (km/h) Year 2 Wind Speed (km/h)

Probability distributions of annual wind speeds for two simulated years

0.35 0.35
0.30 Average = 9.56 MW 0.30 Average = 10.10 MW
Probability

Probability
0.25 Capacity credit = 0.24 0.25 Capacity credit = 0.25
0.20 0.20
0.15 0.15
0.10 0.10
0.05 0.05
0.00 0.00
0 4 8 12 16 20 24 28 32 36 40 0 4 8 12 16 20 24 28 32 36 40
Year 1 Power Output (MW) Year 2 Power Output (MW)

Probability distributions of annual power outputs for two simulated years

39
 0.25

0.2 Observed Wind Speeed

Simulated Wind Speed
Probability

0.15

0.1

0.05

0
0 10 20 30 40 50 60 70 80 90 100
Wind Speed (km/h)
Fig. 1. Observed and simulated wind speed distributions for the Swift Current site

40
 0.35

0.3 Simulat ed Wind Dat a

0.25 Observed Wind Dat a

Probability

0.2

0.15

0.1

0.05

0
0

17.5
22 5
27.5
32 5
37 5
42.5
47 5
52.5
57 5
62.5
67 5
72 5
77.5
82 5
87.5
92 5
97.5
.5
0
5
12 5
2.
7.

10
.

.
.

.
.

.
Capacit y Out age Level (%)

Fig. 2. Capacity outage probability profile for the WTG unit

41
Five State Capacity Outage Probability Table for
a 20 MW WECS

Capacity Probability
Outage
(MW) FOR = 0% FOR = 4%
0 0.07021 0.05908
5 0.05944 0.06335
10 0.11688 0.11475
15 0.24450 0.24408
20 0.50897 0.51875
DAFORW 0.76564 0.77501

42
 LOLE Versus Peak Load

• 5*20 MW (FOR = 0.03) Plus 20 MW wind

20MW Wind
100

LOLE (hours/year)
Multi-state wind model 10
Cap. Out Individual 1
(MW) Probability 0.1
0 0.07021
0.01
5 0.05944 40 45 50 55 60 65 70 75 80
10 0.11688
Peak Load (MW)
15 0.24450
20 0.50897 No Wind 20 MW Wind, FOR=0%

43
 Aleatory and Epistemic Uncertainty
There are two fundamentally different forms of
uncertainty in power system reliability assessment
The component failure and repair processes are random
and create variability known as aleatory uncertainty.
There are also limitations in assessing the actual
parameters of the key elements in a reliability assessment.
This is known as epistemic uncertainty. It is knowledge
based and therefore can be reduced by better information.

It is important to recognize the difference in aleatory and

epistemic uncertainty.

44
 Representation of Load Forecast
Uncertainty
It is difficult to obtain sufficient historical data
to determine the distribution type and the most
common practice is to describe the epistemic
uncertainty by a normal distribution with a
given standard deviation. The distribution
mean is the forecast peak load. The load
uncertainty represented by a normal
distribution can be approximated using the
discrete interval method, or simulated using the
tabulating technique of sampling.

45
 Risk Evaluation with Load Forecast
Uncertainty (LFU)
Assume the load forecast uncertainty 0.8
is represented as in the figure.

Probability
0.6
3
LOLE   Pi * LOLEi 0.4

i 1
0.2
0
Pi is the probability of each load level 55 60 65
Peak Load (MW)
LOLEi is the LOLE for each load level

Peak Load LOLEi Probability C2*C3

(MW) (hrs/year)
55 1.248490 0.2 0.249698
60 1.518238 0.6 0.910943
65 12.816507 0.2 2.563301
1.0 LOLE = 3.723942
46
 Study System-RBTS Data
• Installed Capacity = 240 MW
Unit Type No. of MTTF Failure MTTR Repair FOR
(MW) Units (hr) Rate (hr) Rate
(occ/yr) (/yr)
5 Hydro 2 4380 2.0 45 198 0.010

10 Lignite 1 2190 4.0 45 196 0.020

20 Hydro 4 3650 2.4 55 157 0.015

20 Lignite 1 1752 5.0 45 195 0.025

40 Hydro 1 2920 3.0 60 147 0.020

40 Lignite 2 1460 6.0 45 194 0.030

• Peak Load=185MW
• The load duration curve is taken from the IEEE-RTS
47
 RBTS Analysis at HLI

• Basic System- LOLE versus Peak Load

10
LOLE (hours/year)

0.1

0.01

0.001
135 145 155 165 175 185 195 205
Peak Load (MW)

48
 RBTS Analysis at HLI

• LOLE versus WTG total capacity

• Installed Capacity=240 MW, Peak Load =185MW
North Battleford Saskatoon Regina

1.2

0.8
LOLE(Hour/year)

0.6

0.4

0.2

0
0 10 20 30 40 50 60 70 80 90 100
WTG Total Capacity (MW)
49
 RBTS Analysis at HLI
Add 20 MW wind power to the RBTS

10
LOLE (hours/year)

IPLCC=4.8 MW

0.1
165 175 185 195 205
Peak Load (MW)

Capacity Credit (CC) = IPLCC/Wind Capacity

= 4.8/20.0 = 0.24 = 24%

50
An important consideration in adequacy
evaluation of power systems containing wind
energy is the reliability contribution that WTG
units make compared with that of conventional
generating units.
In order to investigate this, different units in
the reliability test system were removed, and
the number of WTG units required to maintain
the criterion reliability was determined.

51
 System Studies
• Two published reliability test systems with
different capacities, the RBTS and the IEEE
Reliability Test System (IEEE-RTS) were used
in these studies.
• The RBTS consists of 11 conventional
generating units with a total capacity of 240
MW. The total capacity of the IEEE-RTS is
3405 MW. The annual peak load for the RBTS
is 185 MW. The annual peak load is 2850 MW
for the IEEE-RTS.

52
A 5 MW conventional generating unit was first
removed from the RBTS and replaced by WTG units.
A Regina location wind regime was assumed. The risk
criterion is the RBTS original LOLE of 1.05 hours/year.
The LOLE increases from 1.05 hours/year to 1.68
hours/year after the 5 MW unit is removed from the
RBTS. The LOLE is restored to 1.05 hours/year when
45 MW of WTG is added.
This indicates that 45 MW of WTG is able to replace a
5 MW conventional generating unit under this
particular condition. The wind capacity replacement
ratio in this situation is 9.0.

53
 RBTS Analysis at HLI
• Replacement ratio versus mean wind speed
multiplication factor
5 MW Unit Removed 10 MW Unit Removed
20 MW Unit Removed 40 MW Unit Removed
18
16
14
Replacement Ratio

12
10
8
6
4
2
0
1 1.1 1.2 1.3 1.4 1.5
Mean Wind Speed Multiplication Factor
54
Replacement ratio versus mean wind speed
multiplication factor (IEEE-RTS)
12 MW Unit Removed 50 MW Unit Removed
100 MW Unit Removed 350 MW Unit Removed

12

10
Replacement Ratio

0
1 1.1 1.2 1.3 1.4 1.5
Mean Wind Speed Multiplication Factor
55
 Independent Wind Energy Sources

A WTG produces no power in the absence of

sufficient wind and there is a definable
probability that there will be insufficient wind
at a given site.
The probability, however, of there being no
wind simultaneously at two widely separated
independent wind sites is much less, and
locating WTG at independent wind sites can
provide considerable benefits.

56
 Replacement ratio versus the capacity removed
from the RBTS (single, two and three wind farms)
Single Wind Site 2 Wind Sites 3 Wind Sites
Not Available (Single Wind Site) Not Available (2 Wind Sites)

18
Replacement Ratio

16
14
12
10
8
6
4
2
0
5MW Unit 10MW Unit 20MW Unit 40MW Unit
Removed Removed Removed Removed

57
Replacement ratio versus the capacity removed from the
IEEE-RTS (single, two and three wind farms)
Single Wind Site 2 Wind Sites 3 Wind Sites
Not Available (Single Wind Site) Not Available (2 Wind Sites)
Replacement Ratio

12
10
8
6
4
2
0
12MW Unit 50MW Unit 100MW Unit 350MW Unit
Removed Removed Removed Removed

58
 Planning Capacity Credit Evaluation

A sequential Monte Carlo simulation program

developed for generating capacity adequacy evaluation
was used to study the IEEE-RTS at a peak load of 2850
MW. Five 100 MW WECS were added sequentially to
the IEEE-RTS using the Regina wind regime data. The
sampling size for the IEEE-RTS is 20,000 years.

59
Effects on the System Reliability Indices of
Adding Wind Power

The added wind capacity is considered to be either

completely dependent or fully independent. These
conditions may not exist in an actual system and there will
be some degree of cross-correlation between the site wind
regimes. The dependent and independent conditions
provide boundary values that clearly indicate the effects
of site wind speed correlation.

60
 Increase in Peak Load Carrying
Capability with Added Wind Power

The IEEE-RTS IPLCC

as a function of the
added wind capacity

61
The IEEE-RTS Wind Planning Capacity
Credit (PCC) with Sequential Wind
Power Additions Based on LOLE

62
The IEEE-RTS Wind Planning Capacity
Credit (PCC) with Sequential Wind
Power Additions Based on LOEE

63
The IEEE-RTS IPLCC as a function of the
added conventional generating capacity
based on the LOLE and LOEE

64
Security Based Adequacy Evaluation Using
the System Well-Being Approach

The system well-being approach provides a

combined framework that incorporates both
deterministic and probabilistic criteria. The
combination of deterministic and probabilistic
concepts occurs through the definition of the
system operating states.

65
Security Based Adequacy Evaluation Using the System
Well-Being Approach
Healthy state – all equipment and
Success operating constraints are within limits
and there is sufficient margin to serve
Healthy the total load demand even with the
loss of any element (i.e. the N-1
deterministic criterion is satisfied.).
Marginal Marginal state – the system is still
operating within limits, but there is no
longer sufficient margin to satisfy the
At Risk acceptable deterministic criterion.
At risk state – equipment or system
System Well-Being Framework constraints are violated and load may
be curtailed.

66
Security Based Adequacy Success
Evaluation Using the System Healthy
Well-Being Approach
Marginal

At Risk
System Well-Being
Framework

Healthy Marginal
At Risk

Cliff 67
 Security Based Adequacy Evaluation Using
the System Well-Being Approach
Success
Healthy System Well-Being Indices:
Prob{H} Freq{H}
Marginal Prob{M} Freq{M}
Prob{R} Freq{R}
At Risk
System Well-Being Framework

68
 Security Based Adequacy Evaluation Using the
System Well-Being Approach
Base Case – RBTS with no wind generation.
Case A – RBTS with a 10 MW unit replaced by 2 – 18 MW
wind farms at W1 and W2.
Case B – RBTS with a 10 MW unit replaced by 3- 9 MW
wind farms at W1, W2 and W3.
The system P(R) is 0.00043 in all three cases.
Wind Farm W1 W2 W3
Mean Wind Speed (m/s) 9.10 8.38 10.03
Standard Deviation (m/s) 5.50 4.48 5.20
Correlation w.r.t W1 1.00 0.85 0.05

69
 Security Based Adequacy Evaluation Using the
System Well-Being Approach
Index Base Case Case A Case B
P(H) 0.98456 0.98130 0.97834
P(M) 0,01501 0.01827 0.02122
P(R) 0.00043 0.00043 0.00043
F(H) occ./ yr 25.1 33.9 36.3
F(M) occ./ yr 25.8 34.9 37.1
F(R) occ./ yr 0.8 1.0 0.9
D(H) hrs./ occ. 403.2 283.7 263.3
D(M) hrs./ occ. 5.1 4.6 5.01
D(R) hrs./ occ. 4.6 3.6 3.8

70
 Epistemic Uncertainty
Load growth and load forecast uncertainty are affected
by social, political, environmental and economic
factors.

Load forecast uncertainty also depends on the required

length of time in the future of the forecast. Different
types of generating capacity have different lead times
that involve regulatory and environmental approvals.

Nuclear - 8 to 10 years, Hydro - 6 to 8 years,

Fossil - 5 to 6 years, Gas turbines - 2 to 3 years,
Wind -1 to 2 years.

71
 RBTS Analysis at HLI
• Considering Load Forecast Uncertainty

100
LOLE (hrs/year)

10

0.1

0.01
145 155 165 175 185 195 205
Peak Load (MW)

LFU S.D.=0 5% 10%

72
 Aleatory Uncertainty

The Loss of Load (LOL) in a given period is a random

variable and is dependent on the failure and repair
processes of the system components.

The LOLE is the mean value of the LOL distribution.

73
 RBTS Analysis at HLI
• LOL Distribution Peak Load Zero LOL
Relative Frequency % Relative Frequency % Relative Frequency % (MW) Probability
8 E(X) = 1.0901
6 S. D. = 4.2210
185 86.49%
4
2
0
1.5 6.0 10.5 15.0 19.5 24.0 28.5 33.0
8 Loss of Load (hours/year)
E(X) = 2.5135
6 S. D. = 6.6007
4 195 73.70%
2
0
1.5 6.0 10.5 15.0 19.5 24.0 28.5 33.0
8 Loss of Load (hours/year)
E(X) = 5.3243
6 S. D. = 9.9752
4
2
205 52.59%
0
1.5 6.0 10.5 15.0 19.5 24.0 28.5 33.0
Loss of Load (hours/year)
74
Example Reliability Criterion – NERC Region XXX
“Sufficient megawatt generating capacity shall be installed to
ensure that in each year for the XXX system the probability of
occurrence of load exceeding the available generating capacity
shall not be greater, on the average, than one day in ten years.
Among the factors to be considered in the calculation of the
probability are the characteristics of the loads, the probability
of error in load forecast, the scheduled maintenance
requirements for generating units, the forced outage rates of
generating units, limited energy capacity, the effects of
connections to the pools, and network transfer capabilities
within the XXX systems.”

75
 Scheduled Maintenance
Period Evaluation Method

76
 Period Analysis

n = 12 in monthly analysis
= 4 in seasonal analysis

77
• Different reliability indices are obtained using different
load models.

• The LOLE index in hours is obtained using hourly load

values.
• The LOLE index in days is evaluated using daily peak load
values.

• It is not valid to obtain the LOLE in hours by multiplying

the days/year value by 24. The commonly used index of 0.1
days/year, which is often expressed as one day in ten years,
cannot be simply converted to an equivalent index of 2.4
hours/year. This is because the hourly load profile is
normally different from that of the daily peak load.

78
 Load Models
Daily peak load variation curve (DPLVC) – LOLE in days/year

Load duration curve (LDC) – LOLE in hours/year &

energy based indices, UPM

1.0

0.9
Load (p.u of Peak)

0.8

LDC
0.7 DPLVC

0.6

0.5

0.4
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

Time (p.u.)

79
 Basic RBTS HLI Analysis
The following studies were done using two general generating
capacity adequacy evaluation programs.

Reliability Analytical Program Simulation Program

Index
Constant Daily Peak Hourly Constant Daily Peak Hourly
Load Loads Loads Load Loads Loads

LOLE 3.0447 0.1469 - 3.0258 0.1496 -

(days/year)

LOLE 73.0728 - 1.0919 72.6183 - 1.0901

(hours/year)

LOEE 823.2555 - 9.8613 816.8147 - 9.9268

(MWh/year)

LOLF - - - 2.8309 0.2171 0.2290

(occ/year)

80
Ratio of the LOLE (hours/year) over the LOLE
(days/year) for the RBTS

10
8
6
Ratio

4
2
0
125 135 145 155 165 175 185 195 205 215 225 235
Peak Load (MW)

81
Ratio of the LOLE (hours/year) over the LOLE
(days/year) for the IEEE-RTS

8
7.5
Ratio

7
6.5
6
2736 2793 2850 2907 2964 3021 3078 3135 3192 3249
Peak Load(MW)

82
LOLE(hours/year) and LOLE (days/year)
The LOLE in days/year provides a more
pessimistic appraisal than that given by the
LOLE in hours/year. The two test systems have
the same normalized chronological hourly load
model and therefore the same daily and annual
load duration curves. The system load factor is
61.44%. The ratio difference in the two test
systems is therefore due to the different
generation compositions.

83
The reciprocal of the LOLE in years per day is
often misinterpreted as a frequency index. As
an example, the commonly used LOLE index of
0.1 days/year is often expressed as one day in
ten years and extended to mean “once in ten
years”. This is not a valid extension and has a
frequency of load loss connotation that is not
present in the LOLE index. In order to
illustrate this, a comparison of the LOLE
(days/year) and LOLF (occ/year) indices was
conducted using the two test systems.

84
 Ratio of the Reciprocal of the LOLE
(days/year) over the Reciprocal of the LOLF
(occ/year) for the RBTS.

2
1.5
Ratio

1
0.5
0
145 165 185 205 225
Peak Load (MW)

Using Daily Peak Load Using Hourly Load

85
 Ratio of the Reciprocal of the LOLE
(days/year) over the Reciprocal of the LOLF
(occ/year) for the IEEE-RTS.

1.5

1
Ratio

0.5

0
2622 2736 2850 2964 3078
Peak Load (MW)

Using Daily Peak Load Using Hourly Load

86
Reliability Index Probability Distributions

The simulation program was applied to the

IEEE-RTS to create the reliability index
probability distributions.
The load is represented by the hourly values.
The sampling size for the IEEE-RTS is 20,000
sampling years, which provides a coefficient of
variation less than 1%.

87
 LOLE and Probability of Zero LOL for the IEEE-RTS.
Peak Load LOLE LOL Standard Probability of
(MW) (hours/year) Deviation no LOL
2850 9.39 16.49 43.35%
2964 19.36 24.99 21.12%

3078 36.33 35.66 7.04%

LOEE, LOLF and the Standard Deviations for the IEEE-RTS.

Peak Load LOEE LOE Standard LOLF LOLF Standard
(MW) (MWh/year) Deviation (occ/year) Deviation

2850 1192.51 3061.14 2.00 2.79

2964 2621.69 4891.98 3.98 4.06

3078 5214.57 7407.87 7.21 5.59

88
 The Distribution of the LOL for the IEEE-RTS.
16
Relative Frequency %

14
12
10
8 Peak Load = 2850 MW
6
4
2
0 P (zero LOL) = 43.35%
4 16 28 40 52 64 76 88 100 112 280
Loss of Load (hours/year)

16
Relative Frequency %

14
12
10 Peak Load = 2964 MW
8
6
4
2
P (zero LOL) = 21.12%
0
4 16 28 40 52 64 76 88 100 112 296
Loss of Load (hours/year)

16
Relative Frequency %

14
12
10
8 Peak Load = 3078 MW
6
4
2
0
P (zero LOL) = 7.04%
4 16 28 40 52 64 76 88 100 112 304
Loss of Load (hours/year)

89
The Distribution of the LOE for the IEEE-RTS.
30
Relative Frequency %

25
20
15 Peak Load = 2850 MW
10
5
0
250 1000 1750 2500 3250 4000 4750 5500 6250
Loss of Energy (MWh/year)

30
Relative Frequency %

25
20
15 Peak Load = 2964 MW
10
5
0
250 1000 1750 2500 3250 4000 4750 5500 6250
Loss of Energy (MWh/year)

30
Relative Frequency %

25
20
15 Peak Load = 3078 MW
10
5
0
250 1000 1750 2500 3250 4000 4750 5500 6250
Loss of Energy (MWh/year)

90
As noted earlier, the LOLE index is the most
commonly used adequacy index in generating
capacity planning. The LOLE does not contain
any information on the magnitude of load loss
due to insufficient generation. It simply
indicates the expected number of hours of load
loss in a given year. The LOEE is a more
complex index and is a composite of the
frequency, duration and magnitude of load
loss.

91
 The LOEE can be combined with an index known as
the Interrupted Energy Assessment Rate (IEAR) to
give the expected customer economic loss due to
capacity deficiencies. Assuming an IEAR of 15.00/kWh
of unserved energy, the expected customer interruption
costs (ECOST) are as follows:

Peak Load (MW) ECOST($)

2850 17,887,608
2964 39,325,287
3078 78,218,605

These values were obtained by taking the product of the

IEAR and the respective LOEE.
92
Additional information on the likelihood of
encountering a particular level of monetary loss can be
obtained using the distribution in the previous figure.
As an example, the relative frequencies of encountering
a monetary loss exceeding 900 million dollars are as
follows.

Peak Load (MW) Relative Frequencies(%)

2850 5.38
2964 13.28
3078 28.03

The distributions provide considerable additional

information that can be used in electricity utility risk
assessment and management.
93
The Distribution of the LOLF for the IEEE-RTS
16
Relative Frequency %

14
12
10
8
6
4 Peak Load = 2850 MW
2
0
1.0 3.0 5.0 7.0 9.0 11.0 13.0 15.0 17.0 19.0 21.0 23.0
Loss of Load Frequency (occ/year)

16
Relative Frequency %

14
12
10
8
6 Peak Load = 2964 MW
4
2
0
1.0 3.0 5.0 7.0 9.0 11.0 13.0 15.0 17.0 19.0 21.0 33.0
Loss of Load Frequency (occ/year)

16
Relative Frequency %

14
12
10
8
6 Peak Load = 3078 MW
4
2
0
1.0 3.0 5.0 7.0 9.0 11.0 13.0 15.0 17.0 19.0 21.0 42.0
Loss of Load Frequency (occ/year)

94
The basic generating capacity adequacy indices
can be determined using analytical techniques or
simulation methods.
Simulation can be used to provide a wide range of
indices, to incorporate complex operational
constraints, and create reliability index
probability distributions.

95
1. “Reliability Evaluation of Power Systems,
Second Edition”, R. Billinton and R.N. Allan,
Plenum Press, 1996, pp. 514.
2. “Reliability Assessment of Electric Power
Systems Using Monte Carlo Methods”,
R. Billinton and W. Li, Plenum Press, 1994, pp.
351.

96