121.

00-6

Mathematics - Course 121

THE BINOMIAL DISTRIBUTION AND

POWER SYSTEM RELIABILITY

I THE BINOMIAL DISTRIBUTION

Recall (from 121.00-3) that the number of different

combinations of r objects which can be formed from n
different objects is

n!
r!(n-r)!

Suppose that an 'experiment' is tried repeatedly and

that all of the following apply:

1. the total number of trials, n, is fixed.

2. each trial has only two possible outcomes:

'success' with probability p, and

'failure' with probability q,

where p + q = l.

3. p and q are the same for all trials.

4. the outcome of anyone trial is independent of all

others.

Then the probability of exactly r 'successes' is

P = C p r q n-r •
r n r
In this expression, prqn-r represents the probability
(given by PRl) of exactly r 'successes' and n-r 'failures' in
some particular order, eg, r consecutive 'successes' followed
by n-r consecutive 'failures'. The coefficient nCr accounts
for the number of possible ways to order r 'successes' and
n-r 'failures'.

The distribution of probability over the n+l possible

numbers of successes, r = 0, 1, 2, ••• n, is called the
Binomial Distribution. The binomial distribution takes its
name from the binomial expansion,

n
(p + q}n = l:
r n-r
C p q •
n r
r=O

June 1981 - 1 -
 121. 00-6

Example 1

Plot the probability distribution of the number of heads

obtained in 5 tosses of a coin.

Solution

In this case. n = 5, and p = q = 1/2

0 5 1
Po = 5 COP q = IT
4 5
P
1 = sclpq = IT
'2 3 10
P
2 = 5 C2 P q = IT
3 2 10
P
3 = SC 3 P q = 32

4 5
P
4 = SC 4 P q = 32

5 0 1
Ps = SC 5 P q = 32

Pr 12 -.
(x32 )
~

8 -f-

-
4 -
-
o 11
o 1 2 3 4 5

r Heads

Figure 1

Probability Distribution of Number of Heads in 5 Coin Tosses

- 2 -
 121.00-6

Example 2

Using Probability Rule #9 from 121.00-3, prove that the

mean of the binomial distribution is np.

Proof

n
E(r) = L rP (by PR9)
r=O r

n
= L: r n: prqn-r (0 •• no contribution from r=O)
r:(il-r):
r=l

n (n-1)! ( ~ r-l q [(n-l)-(r-l)])

= np L:
r=l

m
m: p s q m-s
= np L:
s: (m-s) :
(m = n - 1)
s=O (s = r - 1)

= np (p + q)m (using binomial expansion)

np (1) m (p + q = 1)

= np QED

Example 3

Plot the probability distribution of successes in 4 trials,

given that p = 1/4. What is the expected number of successes?

Solution

- 3 -
 121.00-6

P
r
(x2 56)

I
o
o 1 2 3 4
r Successes

Figure 2

Probability Distribution of
Successes in 4 Trials

Using the result of Example 2 above, the expected number

of successes, E(r) = np
= 4 x 1/4
1

Example 4

If stock is 99.5% free of defects, what is the expected

number of defective items in a random sample of 20 items?

Solution

Probability of defect in each item examined, p = 0.005.

Then expected number of defects = np
= 20 x 0.005
0.1

Example 5

A system consisting of 8 dousing valves is considered

failed if 3 or more dousing valves have failed. If the
unavailability ot a dousing valve is 0.01, calculate the
unavailability of the system.

- 4 -
 121.00-6

solution

_.
Let Q' \ 1
, !Q
=:
represent
.
unavailabilities
_ _
of a dousing valve,
douslng system~ respectlvely. Then

Qs = P(exactly 3 valves failed) + P(exactly 4 valves failed)

+ ••• + P(exactly 8 valves failed) (by PR4)

= C Q 3R 5 + C Q 4 R 4 +
8 3 v v 8 4 v v

8' 3 5 8' 4 4
= 3!5! (.01) (.99) + 4!4! (.01) (.99) + ••• +

= 5.3 X 10 -5 + 6.7 x 10 -7 + ... + 10 -16

= 5 x 10- 5

Note that only the lead term on the RHS contributed signi-
ficantly to the answer, the probabilities of higher order fail-
ures (failure of 4, 5, ••• , 8 valves) being negligible by
comparison to the probability of 3 valves failing.

II PO~mR SYSTEM RELIABILITY

The average daily variation of grid load is shown

schematically in Figure 3.

Peak Load
Load (MW)

Base Load

o 12 24
Time (Hours)
Figure 3
Average Daily Load Variation

- 5 -
 121.00-6

A utility very seldom gets into trouble supplying its

base load, but is much more likely to encounter difficulty
supplying the peak load. (Hence, the move towards providing
incentives for switching to off-peak consumption.) Figure 4
shows the nu~~cr of days per annum that daily peak load
exceeds any given value. Also shown on Figure 4 is the
installed capacity, the reserve capacity (installed minus
maximum daily peak), and the lost capacity, Ok' associated
wi th the kth poss ible outage, and tk' the number of days per
annum that Ok causes a load loss.

---------} Reserve Capacity

Installed Capacity
----- __________
Daily Peak
Load (MW) Ok

o Days/Year 365

Figure 4

Typical Annual Daily Peak

Load Variation Curve

DEFINITION
The E#pected Load Curtailment (ELC) is the number of
days per annum for which a load loss occurs, ie,

where P k is the probability of the kth possible outage, and

tL
is the number of days/year that the kth outage causes
K
a load loss.

- 6 -
 121.00-6

DEFINITION

The Expected Load Loss (ELL) is the expectation value of

the load loss without regard to time, ie,

[,-_E_L_L_=_~_Pk_Lk
__ l
where Pk is the probability of the kth outage, and

Lk is the load lost on the kth outage.

Both ELC and ELL are used as measures of power system

reliability. A utility would set a reliability target of,
say, <0.1 day/y for the ELC.

DEFIl'~ITIO~J

Forced outage rate is the fraction of time on forced

outage, ie,

Time on forced outage

FOR = Time on forced outage + Time in operation

Note that for simplicity, this lesson assumes that gen-

erating units are either operating at full output or are down
on forced outage. Other possible states such as 'operating
derated' or 'available but not operating' are not considered.

Example 6

A small power system consists of three identical 5 MW

generating units, each with a forced outage rate (FOR) of
0.03. Assuming a continuous load of 10 MW, draw up a capa-
city outage probability and expected load loss distribution
table. Calculate the expected load loss and the expected
load curtailment.

solution

The Annual Daily Peak Load Variation Curve for this

example is shown in Figure 5, and the required capacity out-
age probability and load loss distribution table is given in
Table 1.

- 7 -
 121.00-6

Daily Peak
Load (MW)
15
----------------1
Installed Capacity

Reserve 5 MW

10

o
o Days/Year 365

Figure 5

Annual Daily Peak Load

Variation Curve for Example 6

Note that any outage Ok resulting in a load loss Lk has

an associated tk = 365 days/yo

The ELL and ELC are calculated using Table 1 as follows:

ELL = L LkPk
k
= 0.013365 MW

all losses
ELC = L Pktk
k

= (0.002619 + 0.000027) (8760 hours/y)

= 23 hours/y

- 8 -
 1.21.00-6

Example 6

Table 1

Capacity Outage Probability and Load Loss Distribution Table

Outage Ok

Number Capacity Probability Load Loss

k Units Out Out (MW) Pk Lk (MW) LkPk(MW)

1 0 0 0.912673 0 0

2 1 5 0.084681 0 0

3 2 10 0.002619 5 0.013095

4 3 15 0.000027 10 0.000270

Note ~ Pk = E 3Cr(.97)r(.03)3-r = (.97 + .03)3 = 1.000000

k r

That E Pk = 1 can be used as an internal check when drawing

k up such tables.

Example 7

A power system contains the following generating capa-

city:

3 x 40 MW hydro units, FOR = 0.005

1 x 50 MW thermal unit, FOR = 0.02

1 x 60 MW thermal unit, FOR = 0.02

a) Calculate the ELL and ELC assuming a continuous load of
200 MW.

b) Calculate the ELC assuming the annllal daily peak load

variation curve decreases linearly from 200 MW at t = 0
to 80 MW at t = 365 d/y.

- 9 -
 121.00-6

Solution

a) The complete capacity outage probability distribution

table including load losses for the case of a 200 MW
continuous load is given in Table 2.
From Table 2,

= 1.181975 MW

all losses
ELC = L Pktk
k

16
= (l: P )
k
(365 d/y) (t k = 365 diY, k = 2, •.• ,16)
k=2

= 19.7 days/year

16
(Note that L P
k
= 1 - PI in the above.)
k=2
Example 7(a)

Outage Ok
k Units Capacity Pk 4< ~Lk

1 none 0 (.995)3( .98)2 .946066 0 0

2 40 40 3Cl(·995)2(.005)(.98)2 = .014262 10 .142623
3 50 50 (.995)3(.02)(.98) = .019307 20 .386148
4 60 60 (.995)3(.98)(.02) = .019307 30 .579224
5 2 x 40 80 3C2(·995)(.005)2(.98)2 = .000072 50 .003583
6 40, 50 90 3Cl(·995)2(.005)(.02)(.98) = .000291 60 .017464
7 40, 60 100 3Cl(·995)2(.005)(.98)(.02) = .000291 70 .020375
8 50, 60 110 (.995)3(.02)2 = .000394 80 .03152Z
9 3 x 40 120 3C3(·005)3(.98)2 = .000000 90 .000011
10 2x40,50 130 3CZ(·995)(.005)2(.OZ)(.98) = .000001 100 .000146
11 2x40,60 140 3C2(·99S)(.00S)2(.98)(.02) = .000001 110 .000161
lZ 40,50,60 150 3Cl(·995)2(.005)(.02)Z = .000006 120 .000713
13 3x40,50 170 (.005)3(.02)(.98) = .000000 140 .000000
14 3x40,60 180 (.005)3(.98)(.02) = .000000 150 .000000
15 Zx40,50.60 190 3CZ(·995)(.005)2(.OZ)Z - .000000 160 .000005
16 3x40,50,60 200 (.005)3(.02)2 - .000000 200 .000000
--- ----~- --- - - --- - - -

Table 2

Capacity Outage Prbbability Distribution and Load Loss Table

- 10 -
 l~l.OO-b

b) The annual daily peak load variation curve is shown in

Figure 6.

240 Installed Capacity = 230 MW

Daily Peak
Load (MW)

120
----- -;:-
------ ---------~ Reserve Capacity

J~---
= 30 MW

80

0--------------1-
o Days/Year 365

Figure 6

Annual Daily Peak Load Variation Curve for Example 7(b)

By similar triangles in Figure 6,

L 120 MW
k
tk
= 365 d/y

365
ie, t
k = 120
L
k

The complete capacity probability and load loss table for

Example 7(b) is given in Table 3, from which

16
ELC = L: Pkt k
k=2

= 3.6 days/year

- 11 -
 121. 00-6

Example 7 shows that, even with a few units, the capacity

outage probability distribution and load loss table can become
complex. Examination of Tables 2 and 3 shows, however, that
there are only a few outages which contribute significantly to
the expected load curtailment. In practice, the table is truncated
at some specified cumulative probability, say, for example, at
99.9999%. This truncation might cut out hundreds or even thous-
ands of entries in the table for a large power system.

Example 7(b)

Otc 365
k Units Capacity 11t Lk tk • 120 ~ l'Jt tk
1 0 0 .946066 0 0 0
2 1 x 40 40 .014262 10 30.4167 .433803
3 50 50 .019307 20 60.8333 1.174509
4 60 60 .019307 30 91.2500 1.761764
5 "I. x 40 80 .000072 50 152.0833 .010950
6 40, 50 90 .000291 60 182.5000 .053108
7 40, 60 100 .000291 70 212.9167 .061959
8 50, 60 110 .000394 80 243.3333 .095873
9 3 x 40 120 .000000 90 273.7500 .000033
10 2x40,50 130 .000001 100 304.1667 .000304
11 2x40,60 140 .000601 110 334.5833 .000335
12 40,50,60 150 .000006 120 365 .002190
13 3x40 ,50 170 .000000 140 365 .000001
14 3x40,60 180 .000000 150 365 .000001
15 2x40,50,60 190 .000000 160 365 .000004
16 3x40,50,60 230 .000000 200 365 .000000

Table 3

Capacity Outage Probability Distribution and Load Loss Table

- 12 -
 121. 00-6

ASSIGNMENTS

1. A dousing system consisting of 14 identical dousing

valves is considered failed if 3 or more valves are
failed. If valve availability is 0.99, calculate the
system unavailability, and compare with that of Example
5 in the text.

2. A power system has two generating units:

A with capacity 50 MW and availability 0.95, and

B with capacity 60 MW and availability 0.96.

The annual load variation curve for the system is shown

below:

120 Installed Capacity 110 MW

Load (MW) ----------------------- .... -

80
(.2,55)

40

(1.0,20)

O'----+--+---+--t----+-......--+-__....---+--4--
o .5
Year/Year Load Exceeded
1.0

a) Draw up a capacity outage probability distribution

table and calculate the expected load curtailment
for the system.

b) A 10 MW generator C with availability 0.96 is added

to the system. Recalculate the ELC and explain why
such a relatively small change in overall capacity
effects such a large reduction in the ELC.

- 13 -
 121.00-6

3. Par each of the following transformer systems, calculate

and compare

a) The expected percentage load curtailment.

b) The expected load curtailment in hours/year.

Assume all transformers have a FOR of 0.015.

i) 3 transformers each rated at 100% of full load.

ii) 3 transformers each rated at 90% of full load.

iii) 3 transformers each rated at 50% of full load.

iv) 4 transformers each rated at 33 1/3% of full load.

L. Haacke

- 14 -