Вы находитесь на странице: 1из 74

Elec7309

Power System Planning & Reliability


Generating Capacity Adequacy
Olav Krause

Outline
Generation capacity adequacy evaluation
Individual element model and combined state space
Simple two state model
Four state model

Capacity outage probability (COP) table


Binomial expansion technique for homogeneous states
Recursive algorithm for unit removal next lecture

Load Duration Curve (LDC)


Indices
LOLE
EENS

(Loss of Load Expectation)


(Expected Energy Not Supplied)

PLCC

(Peak Load Carrying Capability

Textbook chapter 2

CRICOS Provider No 00025B

Generating Capacity Adequacy


Based on the purposes of adequacy study:
Static capacity
Required during the planning process / for long term evaluation of system
requirement

Operating capacity
To evaluate the operational capacity / actual capacity required to meet the demand
for short term

Based on the techniques of adequacy study:


Deterministic
Percentage reserve margin (%RM), loss of largest unit/contingency, USE % etc.

Probabilistic
Loss of load probability or expectation and loss of energy expectation etc.

Disadvantages in the use of % RM


It tries to compare the capacity requirement based on the system
peak load experienced
If the demand characteristics, type and sizes of installed or planned
generating capacity are different, large differences in capacity
requirement may be observed in order to fulfill the same level of
reliability for two different system with same peak load
No penalty to the large generator unit size (unless it exceeds %RM)

Probabilistic methods
Provides an analytical basis for capacity planning that covers the
following effects:

Partial or complete integration of systems


Capacity of interconnections
Effects of unit size and design
Effects of maintenance schedules and
Other system parameters

Study based on the probabilistic techniques can compare economic


aspects associated with the reliability.

Basic risk (adequacy) eval. model


Generation
model

Load model

Risk model

System representation for generation capacity adequacy evaluation


Total Generation

Total load

Based on aggregated models with no consideration of the


interconnecting electricity network

Generation Adequacy Assessment

MODEL AGGREGATION
CRICOS Provider No 00025B

Generation unit unavailability


1. Simple two state model:
Applicable for base load plant with relatively long operating cycle
Basic parameter used in static capacity evaluation

is the expected forced outage rate (FOR) (#failures/time)


is the expected repair rate (RP) (#repairs/time)

Representation:

Up

Down

Token approach. Only one token which can only reside in one
state at a time Number of state transitions has to be equal

(Un-)availability and failure/repair rates


In order to perform a transition, the originating state has to
be occupied
Mean time of state occupation is equal to state probability

Up

Down

# of expected failures:
# of expected repairs:

CRICOS Provider No 00025B

(Un-)availability and MTTF/MTBR


MTTF (Mean time to failure) m and MTBR (Mean time to
repair) r describe a full operation/repair cycle
In average share of time spent in Up and Down state
equal the ratio in one operation/repair cycle
Duration of total operation/repair cycle:
Average share of time spent in Up state:
Average share of time spent in Down state:
Mean time between failures (MTBF):
CRICOS Provider No 00025B

Simple two state model (contd..)


Unavailabi lity ( FOR)= U =

Here:
= expected failure rate

r
r
f
= =
m+r T

[downtime ]

=
[downtime] + [uptime]

= expected repair rate


m = mean time to failure = MTTF
r = mean time to repair = MTTR
m + r = mean time between failure
= MTBF = 1/f
f = cycle frequency = 1/T
T = cycle time = 1/f

m
m f
Availabili ty= A =
=
= =
+ m+r T

[Uptime]

=
[downtime] + [uptime]

Gen. unit unavailability (contd..)


2. Four state model:
Simple two state model does not incorporate the following:
Scheduled outages
Adequate estimate when the demand cycle is relatively short for example
peaking and intermittent units
Start up and shut downs
IEEE sub committee on probability methods proposed four state model in order to
incorporate above mentioned factors

Four state model (contd..)


(1-PS)/T
Reserve
shutdown

In service
1/D

Here;

T = Average reserve shut down


time between periods of
need

PS/T

Forced out but not


needed

1/T

Forced out
while needed

1/D

D = Average in service time per


occasion of demand
PS = Probability of starting
failure

ELEC7309

CAPACITY OUTAGE
PROBABILITY (COP) TABLE
CRICOS Provider No 00025B

Limiting complexity in ELEC7309


We will in general restrict ourselves to two-state models
We will frequently make use of homogeneous generations
parks, but will have to consider the math behind
inhomogeneous ones
Homogeneous generation parks will have a binomial
distribution, which can be calculated relatively easily

CRICOS Provider No 00025B

Example two generators


Two generators (20MW each):
1
U1

A1
1
2

U2

A2
2

What are the possible state cobinations?

CRICOS Provider No 00025B

State space two generators


No simultaneous failures considered
2
A1,A2

A1,U2

2
U1,A2

U1,U2
2

If both generators are of same capacity, do we really need


all four states? not if were not considering the network
CRICOS Provider No 00025B

State space two generators


What is the resulting total available generation capacity?
2

40MW

A1,A2

20MW

A1,U2
2

20MW

U1,A2

U1,U2
2

Equal for our purpose

CRICOS Provider No 00025B

0MW

State space two generators


40MW

A1,A2

20MW

CRICOS Provider No 00025B

U1,A2

0MW

A1,U2

U1,U2

State space two generators


If both generators are absolutely
identical:

State probabilities become:

Hint: Doesnt that look like this?


CRICOS Provider No 00025B

For small problems - pascal triangle


(p+q)n
n

(p+q)0 0

(p+q)1 1

(p+q)2 2
.
.
.

(p+q)5 5

1
1

=1.p2q0+2.p1q1+1.p0q2

1
3

6
10

=1.p1q0+1.p0q1

1
2

4
5

=1.p0q0

1
4

10

.
.
.

1
6

(p+q)5=1.p5q0+5.p4q1+10.p3q2+10.p2q3+5.p1q4+1.p0q5
CRICOS Provider No 00025B

COP of homogeneous gen. park


Back to our example
This time 5 identical generators with 20MW each

CRICOS Provider No 00025B

Two generator example


5 generators of 20MW with

With this nomenclature,


this is a Capacity Outage
Probability Table!

Failure rate = 3[f/yr]


Repair rate = 97[r/yr]
A=0.97, U=0.03

States and probabilities

100MW
80MW
60MW
40MW
20MW
0MW

CRICOS Provider No 00025B

1A5
5A4U1
10A3U2
10A2U3
5A1U4
1U5

= 0.858734
= 0.132795
= 0.008214
= 0.000254
= 0.000004
= 2.43E-8

Outage:
0MW
20MW
40MW
60MW
80MW
100MW

Two generator example


So, if we assume a constant load of 60MW, what is the
probability well not be able to supply it?
COP table

100MW
80MW
60MW
40MW
20MW
0MW

CRICOS Provider No 00025B

1A5
5A4U1
10A3U2
10A2U3
5A1U4
1U5

= 0.858734
= 0.132795
= 0.008214
= 0.000254
= 0.000004
= 2.43E-8

Outage:
0MW
20MW
40MW
60MW
80MW
100MW

All these states


represent sufficient
available generation
capacity

Two generator example


LOLE(60MW)=0.000257

so, what does that mean?

For 0.0257% of the time there will be some kind of outage


to some unknown extend.
It does not tell:
How frequent and outage events has to be expected
How long such event would on average last
Which likely extend the outage has

It does only tell:


We have to expect some kind of outage event for 0.0257% of the
time (e.g. 0.000257*8760h = 2.25132 hours/year)
CRICOS Provider No 00025B

Cap. outage probability (COP) table


Remark to COP tables:
All the generator units in power system are not identical, so the
binomial distribution cannot always be applied
In such case a recursive technique is used where units are combined sequentially
using basic probability concepts to form a capacity outage table
Theoretically, capacity outage table incorporates all the system capacity
But some states with insignificant probability can be truncated in the capacity
outage probability table
Units with identical capacity can be grouped into the same category prior to
capacity table building

We will look into this a little deeper in the next lecture

CRICOS Provider No 00025B

ELEC7309

LOSS OF LOAD INDICES


CRICOS Provider No 00025B

Aspects we are looking into today


Loss of Load Expectation (LOLE)
LOLE sensitivity
Peak Load Carrying Capability (PLCC)
Effects of capacity extension, scheduled outages and
Load forecast uncertainty
Loss of Energy Expectation (LOEE)/Expected Energy Not
Supplied (EENS)
CRICOS Provider No 00025B

Basic risk (adequacy) eval. model


Generation
model

Load model

Risk model

System representation for generation capacity adequacy evaluation


Total Generation

Total load

Based on aggregated models with no consideration of the


interconnecting electricity network

Load curve

7500

7500
7000
6500
6000
5500
5000
4500
4000

7000
6500
6000
5500
5000

4000

0:00
1:30
3:00
4:30
6:00
7:30
9:00
10:30
12:00
13:30
15:00
16:30
18:00
19:30
21:00
22:30

4500

0:00:00
11:00:00
22:00:00
9:00:00
20:00:00
7:00:00
18:00:00
5:00:00
16:00:00
3:00:00
14:00:00
1:00:00
12:00:00
23:00:00
10:00:00
21:00:00
8:00:00
19:00:00
6:00:00
17:00:00
4:00:00
15:00:00
2:00:00
13:00:00
0:00:00
11:00:00
22:00:00
9:00:00
20:00:00
7:00:00
18:00:00

Demand changes over 0me.


Varia0on during a day (valley, o-peak and peak hours)
Varia0on during week and seasons (winter peak & summer peak)

Queensland Demand 21/07/2011 (30 minutes data) Source: AEMO

Queensland Demand 4/07/2011 18/07/2011 (30 minutes data) Source: AEMO

Load Duration Curve (LDC)


Annual load duraBon curve determines the number of hours per year when the load is
greater or lesser than a given amount.

Load duraBon curve can be calculated for any period of Bme (weekly, monthly, ).


We can plot load duraBon curve by using available load curve data.

CRICOS Provider No 00025B

How to calculate LDC


6000

Load (kW)

Hours in a day

% 0me

5000

5000

12.5

4000

3000

6+3=9

37.5

3000

2000

6+3+5=14

58.33

1000

6+3+5+4=18

75

500

6+3+5+4+6=24

100

2000
1000
0
1 2 3 4 5 6 7 8 9 10111213141516171819202122232425

Load (kW)
Time
(hours)

CRICOS Provider No 00025B

1000

5000

500

3000

2000

Obtaining LDC from load curve

Load Curve

tj=tj1+tj
2

CRICOS Provider No 00025B

Load Duration Curve

LOLE Calculation
Example
A. Load data:
Daily peak load (MW)
Days of occurrences (Days)

57
12

52
83

46
107

41
116

34
47

B. Generation data:
Capacity model from Example of previous slides (two state model 5 units with
20MW each = total installed capacity of 100 MW)

CRICOS Provider No 00025B

LOLE Calculation (contd..)


A. Load data:
Daily peak load (MW)
Days of occurrences (Days)

57
12

52
83

46
107

41
116

34
47

100 MW

43 MW = P(43) = 0.020392

48 MW = P(48) = 0.020392

54 MW = P(54) = 0.000792

57 MW

59 MW = P(59) = 0.000792
66 MW = P(66) = 0.000792

52 MW
46 MW
41 MW
34 MW

12

83

116

107
365 Days

CRICOS Provider No 00025B

47

Loss of Load Expectation

COP table
Available cap.
p(100MW)
p(80MW)
p(60MW)
p(40MW)

p(20MW)
p(0MW)

Ind. probability
= 0.858734
= 0.132795
= 0.008214
= 0.000254

= 0.000004
= 2.43E-8

Load data
p(57MW) = 12/365 = 0.005479
p(52MW) = 83/365 = 0.227397
p(46MW) = 107/365 = 0.293151
p(41MW) = 116/365 = 0.317808
p(34MW) = 47/365 = 0.128767

Equivalent to ~113 minutes per year


CRICOS Provider No 00025B

LOLE Calculation (contd)


Using the above LOLE equation:
LOLE = 12 x P(100-57) + 83 x P(100-52) + 107 x P(100-46) + 116 x P(100-41)+ 47
x P(100-34)
The above formulation gives LOLE in days per year

LOLE = (probability of existence of particular capacity outage x


number of time units (frequently used: days) in the period where loss
of load would occur due to that outage)

CRICOS Provider No 00025B

LOLE Calculation (contd)


A particular capacity outage in excess of the reserve will contribute to
the system LOLE
Capacity outage less than the reserve will not contribute to system
LOLE
Installed Capacity

Daily peak load MW

Qk

tk

Time load exceeds the indicated value


CRICOS Provider No 00025B

Reserve

ELEC7309

LOLE PEAK LOAD


SENSITIVITY AND PLCC
CRICOS Provider No 00025B

LOLE sensitivity analysis


For a given capacity composition, the value of LOLE is dependent on the
system peak load (Table 2.3C)

CRICOS Provider No 00025B

LOLE sensitivity analysis (contd..)


For a given capacity composition and forecasted load, the value of LOLE
depends upon the unit FOR (Table 2.3D)

CRICOS Provider No 00025B

Affect of system peak load on LOLE


Typical shape of LOLE vs.
Peak Demand:
why this step?
At certain demand levels there
are steps in how many units
may fail.
E.g. at 160MW three may fail
as for 180MW a maximum of
two may fail.
This leads to distinct steps in
the LOLE vs. Peak Demand
curve
CRICOS Provider No 00025B

LOLE sensitivity analysis (contd..)


For a small system, effect of FOR on LOLE is pronounced
For larger system, this effect can be pronounced only due to FOR of
larger units
In a large system with the wide range of unit sizes, there will be a continues
capacity outage probability table which results smoother risks profile
This is more pronounced by the addition of a larger units

By investing in design, construction, reliability, maintainability and


spare parts, unit FOR will improve
Worth of this improvement can be appraised based on the
improvement in system risks
To measure this improvement in system Peak Load Carrying
Capacity (PLCC) index is used which will be equivalent to the
investment in new capacity.
CRICOS Provider No 00025B

Peak load carrying capacity (PLCC)


Example:
A large system with 10100 MW capacity and 300 and 500 MW units
The resulting profile for LOLE = 0.1 days/year is a straight line (Figure
2.8)
PLCC shows the impact of unit FOR in terms of system availability
In the system shown in Figure 2.8 and at a risk level of 0.1 days/year:
With FOR = 0.04, PLCC = 9006 MW
With FOR = 0.13, PLCC = 8191 MW

CRICOS Provider No 00025B

Affect of FOR on system PLCC


PLCC is the maximum allowable peak load for which the
system doesnt exceed a predefined LOLE value

CRICOS Provider No 00025B

Peak load carrying capacity (PLCC)

CRICOS Provider No 00025B

ELEC7309

EFFECTS OF CAPACITY
EXTENSION
CRICOS Provider No 00025B

Capacity expansion
Generation expansion planning requires well advance (5 to 10 years)
knowledge of system requirements
Load forecast could be uncertain and it should be taken as random
variable
Numerical example:
Generator data: 5 40 MW and FOR = 0.01
Load data: varying 100-40% with160 MW peak load, 10% increase/year

CRICOS Provider No 00025B

Capacity expansion
Question:
In order to meet the criteria of LOLE 0.15 days/year what year a unit of 50 MW
need to be committed in the above system?
Table 2.17 shows the change in risk level by addition of 50 MW units
Figure 2.11 shows the timing of unit addition
Table 2.19 shows the generation expansion results
However, reliability criteria is only one of the criteria in planning which is studied by
the system administrator
There are many other criteria such as financial and risk assessment, system
stability and transmission constraints etc for any such decision.

CRICOS Provider No 00025B

Capacity expansion

CRICOS Provider No 00025B

Capacity expansion

CRICOS Provider No 00025B

Perturbation effect
Among the probabilistic techniques, LOLE index is also used to look
at the impact of unit size, FOR and load character on reliability
Numerical example:
System A: 5 40 MW, FOR = 0.01 with LOLE criteria 0.1 days/year and added 50
MW unit (Table 2.20)
System B: 10 20 MW, FOR = 0.01 with LOLE criteria 0.1 days/year and added 50
MW unit (Table 2.21)

From loss of largest unit criteria, system A can carry peak load of 160
MW while system B can carry 190 MW
From Table 2.20 and 2.21 (in next slide), we can see that system A
and B are considerably different, a system with smaller units can carry
much higher peak load.

CRICOS Provider No 00025B

Perturbation effect

CRICOS Provider No 00025B

Perturbation effect
Conclusions:
Larger unit addition could be economical from the investment
(economy of scale) and operation cost (better heat rate), but addition
of larger unit will adversely impact on system reliability, this is known
as perturbation effect
From the above tables, impact due to addition of same unit sizes in
this two systems are found to be different
A larger unit addition to the system A (which is composed of smaller
units) will have adverse effect (PLCC penalty) at the beginning and it
will diminish only after certain level of such unit additions and
During expansion planning process, such perturbation effect need to
be carefully examined.

CRICOS Provider No 00025B

ELEC7309

EXPECTED ENERGY NOT


SUPPLIED (EENS)
CRICOS Provider No 00025B

Expect. Energy Not Suppl. (EENS)

Expected energy curtailment concept can also be used to determine the


expected energy produced by each unit in a simple probabilistic production
cost modeling
Generation data (Table 2.4) :
Unit No
1

Load (MW)

75

3
52.5

Capacity (MM)
0
15
25
0
30
0
20

Probability
0.05
0.30
0.65
0.03
0.97
0.04
0.96

30
25

20

Duration (Hours)

100

Load data represented by LDC of 100 Hrs of duration:


CRICOS Provider No 00025B

Expect. Energy Not Suppl. (EENS)

Assume that the economic loading order is unit 1, 2 and 3


Total energy demand =area under the LDC = 4575 MWh
When non of the generator in the system, EENS0 = 4575 MWh
If the system contained only Unit-1, the EENS can be calculated as
shown in table 2.5

CRICOS Provider No 00025B

Expect. Energy Not Suppl. (EENS)


Contribution from Unit 2 can be obtained by adding Unit 2 to the
capacity model of Table 2.5 as shown in Table 2.6

CRICOS Provider No 00025B

Expect. Energy Not Suppl. (EENS)


Final capacity outage probability table of all three is shown in Table
2.7
Expected energy not supplied in the system is 64.08 MWh
Energy Index of Reliability EIR = 1- 64.08/4575.0 = 0.985993

CRICOS Provider No 00025B

ELEC7309

LOAD FORECAST
UNCERTAINTY
CRICOS Provider No 00025B

Load forecast uncertainty


Load forecast uncertainty is extremely important parameter in
evaluating the system reliability
In reality, some uncertainties exist in the forecasted load and this will
affect on the value of LOLE
LOLE increases as the uncertainties in forecast increases
Uncertainties on load can be described by the probability distribution
based on the historical load data
From the experiences, load uncertainties are best described by
normal distribution

CRICOS Provider No 00025B

Load forecast uncertainty


Load uncertainties can be included in risk calculations by two
methods:
Method 1:
Divide the load forecast probability distributions into certain class
intervals and LOLE is computed for each intervals based on the
probability of such load to exist, and sum of them represents the
system LOLE
Method 2:
Modify the load characteristics to represent the uncertainties. The
modified load characteristics is represented by a single load profile.
System LOLE is computed by combining the modified load model with
capacity outage table
CRICOS Provider No 00025B

Load forecast uncertainty


Method 1:
Number of intervals or bins depends on the accuracy desired (Figure
2.15), however the error due to this will be very less
The area of each bin represents the probability of the load being in the
bin with bins mid value
LOLE is computed for each load represented by the 7 bins (Table
2.23)
The sum of the LOLE for each bins is the system LOLE

CRICOS Provider No 00025B

Load forecast uncertainty


Numerical Example:
Generation model: 12 5 MW, FOR = 0.01
Load model: peak 50 MW, 2% std. deviation (uncertainties), load
varies 100-40%
2 % Std. deviation = 50 * 2 /100 = 1 MW (Table 2.23)
Time period: 30 days = 720 Hours
The capacity model is shown in Table 2.22
LOLE results are shown in Table 2.23
Results:
With no load uncertainty, LOLE = 0.25240 (hours/month)
With 2% uncertainties, LOLE = 0.07839425 (hours/month)

CRICOS Provider No 00025B

Load forecast uncertainty

CRICOS Provider No 00025B

Load forecast uncertainty

CRICOS Provider No 00025B

Load forecast uncertainty

CRICOS Provider No 00025B

Load forecast uncertainty


Method 2 (Numerical example):
Equation for straight line 70% Load factor load model can be
represented by; t = 10/6 * (1 - x), Where x = X/L and if X = load in MW,
L = peak load
Load forecast uncertainties are represented by 7 steps approximation
to normal distribution
Conditional load shapes are derived from the following time duration
of load existence and corresponding probability:

t1 = 10/6(1-X/47) for 0 X 47 and probability 0.006


t2 = 10/6(1-X/48) for 0 X 48 and probability 0.061
t3= 10/6(1-X/49) for 0 X 49 and probability 0.242
t4= 10/6(1-X/50) for 0 X 50 and probability 0.382
t5=1 0/6(1-X/51) for 0 X 51 and probability 0.242
t6= 10/6(1-X/52) for 0 X 52 and probability 0.061
t7= 10/6(1-X/53) for 0 X 53 and probability 0.006

CRICOS Provider No 00025B

Load forecast uncertainty


The modified load duration curve is formed by combining the different
segment that are evaluated as shown below:
Segment 1: t =1
for 0 X 18.8
Segment 2: t = 0.006t1 + 0.061t2 + 0.242t3 + 0.382t4 + 0.242t5 + 0.061t6+ 0.006t7
for 18.8 X 47
Segment 3: t = 0.061t2+ 0.242t3 + 0.382t4 + 0.242t5+ 0.061t6 + 0.006t7
for 47 X 48
Segment 4: t = 0.242t3 + 0.382t4 + 0.242t5 + 0.061t6 + 0.006t7
for 48 X 49
Segment 5: t = 0.382t4 + 0.242t5 + 0.061t6+ 0.006t7
for 49 X 50
Segment 6: t = 0.242t5 + 0.061t6 + 0.006t7
for 50 X 51
Segment 7: t = 0.061t6 + 0.006t7
for 51 X 52
Segment 8: t = 0.006t7
for 52 X 53

By calculating t1, t2, t3 etc for given Gen. Capacity In (which is X MW) we
calculate t (time in p.u.) for each segment (results are shown in Table 2.24)
Utilising these t (p.u.), a modified load duration curve (Figure 2.18) is built and
Concept of modified load duration curve is utilised in the range of studies.

CRICOS Provider No 00025B

Load forecast uncertainty

CRICOS Provider No 00025B

Load forecast uncertainty

CRICOS Provider No 00025B

Load forecast uncertainty

CRICOS Provider No 00025B

Whats next?
In class solution of Tut 2

CRICOS Provider No 00025B

Вам также может понравиться