Power Generation Operation Probabilistic production

and Planning, EG2200 cost simulation (PPC)

Probabilistic production cost
simulation of electricity
markets – L12
Lennart Söder
Professor in Electric Power Systems
• Development started in
the late 1960:th
• Analytical calculations
• A fast method but
limitations concerning
possibilities to include
market details.

1 2

Calculation of system index

PPC model
- Basic idea
Assume • Assume a system where all power plants
• Perfect competition are 100 % available.
• Perfect information • In this system it is easy to calculate the
• Load is not price sensitive system indeces from the load duration
curve, LDC
• Neglect grid losses and
limitations
• All scenario parameters can be Example
treated as independent • Load duration curve, LDC
• Two power stations (300 MW, always
Some of these assumption can be avialable, incremental operation cost ß1
treated with some specific methods. and ß2 respectively, ß1  ß2)

3 4
Calculation of system index Calculation of system index
- Basic idea - Basic idea
LOLP = LDC(600) = 876 h/year = 10%

h/year LDC h/year LDC

5 6

Calculation of system index

Load model
- Basic idea
• We only study the total load in the system
• We assume that the load is price
Perfect competition  Unit 1 is used first, since ß1  ß2 independent
h/year LDC
• The load is represented by a
scenarioparameter, D, which has a
probability distribution which is data for the
simulations
• In probabilistic simulation (PPC) the load is
represented with the load duration curve.

7 8
 Load model Load model – Alternative 2

• How to determine the load duration curve? • Definition 6.11. The load curve, D(k),
• It can not be calculated but it has to be states the mean load per hour during a
estimated from historical data and specified time period: k = 1, …, T.
forecasts. D(k) = load hour k [MWh/h]

• Alternative 1: Select a standardized • Definition 6.12. The real load duration

function (e.g. normal distribution) and fit curve, LDCR(k), states the load level which
historical data + forecast to this one. is exceeded during k hours.
• Alternative 2: Calculate an LDC directly LDCR(K) = load level that is exceeded
from available data + forecasts. during k hours [MWh/h]

9 10

Load model – Alternative 2 Load model – Alternative 2

Example 6.7 Example 6.7 Information is

lost when an
LDC is created!

day day

The expected value of the load corresponds to the area below the curve The area below the real LDC is as large as the area below the load curve

11 12
 Load model – Alternative 2 Load model – Alternative 2

• Definition 6.13. The inverted load duration Example 6.8:

curve, LDC(x), states how many hours a
certain load level x is exceeded.
LDC(x)=number of hours when the load
level x is exceeded [h]
By dividing LDC(x) with the length of the
studied time period, the normalized LDC is
obtained.
day
The normalized LDC shows the probability
distribution of the load during the studied
period.
The area below the curve is still the same since only axes are changed

13 14

Load model – Alternative 2 Load model – Practical aspect

Example 6.8:
• In reality the load is a continous stochastic
variable
• To be able to calculate expected energy
values, it is necessary to integrate the load
duration curve, which means that some
numerical methods have to be applied
• It is therefore suitable to use a discrete
approximation of the LDC.

The area is now changed since the y-axes is divided with T. To get the
correct expected value, the area has to be multiplied with T.

15 16
Load model – Practical aspect Thermal power station model
The model of the thermal power stations
include:
• Installed capacity,
• Production cost,
This means that we assume that the
incremental production cost is independent
of the production level, i.e., constant
efficiency.
• Availability,

17 18

Thermal power station model Thermal power station model

• A thermal power station is represented by
a scenario parameter, (available
production capacity), with a distribution
which is used as data for the simulation

• In probabilistic simulation a two state

model is used for available capacity.
• The availability of a thermal power station
can not be calculated, but has to be
estimated from historical data and
forecasts of the future.

19 20
Thermal power station model
- Estimation of availability
Definition 6.14. The Mean Time To Failure
is calculated by
where K is the number of periods when the
power plant is available and tu(k) is the
duration of each of these periods.
21
Thermal power station model
- Estimation of availability
Definition 6.16. The failure rate λ is the
probability that an available unit will fail.
The failure rate can be estimated as
23
Thermal power station model
- Estimation of availability
Definition 6.15. The Mean Time To Repair
is calculated by
where K is the number of periods when the
power plant is not available and td(k) is the
duration of each of these periods.
22
Thermal power station model
- Estimation of availability
Definition 6.17. The repair rate μ is the
probability that an unavailable unit will be
repaired. The repair rate can be estimated
as
24
Thermal power station model
- Estimation of availability
Definition 6.18. The availability is the
probability that a power plant is available.
This probability can be estimated as the
part of a longer period that the unit is
available:
25
Thermal power station model
- Estimation of availability
Example 6.11 (availability in a power
plant). Table 6.4 shows the operation log
of a power plant. Calculate the failure rate,
repair rate and unavailability of this unit.
27
Thermal power station model
- Estimation of availability
Definition 6.19. The unavailability is the
probability that a power plant is
unavailable, which can be estimated by
26
Thermal power station model
- Practical considerations
• How the availability is estimated depends
on available data. Is, e.g., MTTF and MTTR
available or not?
• Remember that availability is not the
same as utilization! If a power plant is
available, does not mean that it is used! A
plant is only used when it is available and
needed! Needed means that the load is
high enough and the plant is competitive.
28
 Power Generation Operation Probabilistic production cost
and Planning, EG2200 simulation (PPC)

• Development started in the late

Probabilistic production cost
simulation of electricity
markets – lecture 13
Lennart Söder
Professor in Electric Power Systems
1960:th
• Analytical calculations
• A fast method but limitations
concerning possibilities to include
market details.

1 2

Calculation of system index

PPC model
- Basic idea
Assume • Assume a system where all power plants
• Perfect competition are 100 % available.
• Perfect information • In this system it is easy to calculate the
• Load is not price sensitive system indeces from the load duration
curve, LDC
• Neglect grid losses and
limitations
• All scenario parameters can be Example
treated as independent • Load duration curve, LDC
• Two power stations (300 MW, always
Some of these assumption can be avialable, incremental operation cost ß1
treated with some specific methods. and ß2 respectively, ß1  ß2)

3 4
Calculation of system index Calculation of system index
- Basic idea - Basic idea
LOLP = LDC(600) = 876 h/year = 10%

h/year LDC h/year LDC

5 6

Calculation of system index

Load model – Alternative 2
- Basic idea
Example 6.8:

Perfect competition  Unit 1 is used first, since ß1  ß2

h/year LDC

The area is now changed since the y-axes is divided with T. To get the
correct expected value, the area has to be multiplied with T.

7 8
Thermal power station model Equivalent load
• We have seen that it is simple to calculate
the system indices in a system where all
power plants are 100 % available
• How is this performed when this is not the
case?
• PPC-method: Concider outages in power
stations as load increase instead of reduced
• The availability of a thermal power station available production capacity.
can not be calculated, but has to be
estimated from historical data and
forecasts of the future.

9 10

Equivalent load Equivalent load - example

Definition 6.10. The equivalent load is A capacity deficit problem (LOLO) occurs when:
given by
Available capacity < demand

Installed capacity – outages < demand

where
Installed capacity < demand + outages
Eg = equivalent load for the power plant
next to be dispatched after unit g, 
D = actual load, Installed capacity < equivalent load
Ok = outage in unit k.

11 12
Equivalent load - example Calculation of system indices
• From the equivalent load duration curve,
it is possible to calculate the system indices
• Total available capacity:

• Total installed capacity:

• Total outage:

13 14

Calculation of system indices Calculation of system indices

• Risk of capacity deficit = Loss of load • Expected Energy Not Served, EENS: This
probability. Capacity deficit occurs when means energy that cannot be delivered
the deman exceeds available production depending on capacity deficit, i.e.,
capacity, i.e. equivalent load > installed capacity

15 16
 Calculation of system indices
• Expected energy production in power
plants, EG: Assume that we do not have
unit g. Then the expected unserved energy
is . When unit g is then installed
the expected unserved energy decreases to
The difference is expected energy
production in the unit.
Calculation of system indices
• Expected production cost. If energy
production in each unit is available, then it
is eacy to calculate the production cost.
• It is also possible to include the cost of
disconnected consumers

17 18

Calculation of system indices Calculation of system indices

• Example 6.3-6.6. In a • Example 6.4. Study the same system as in
simple system the load is example 6.3. What does the equivalent load
constantly 200 kW. The duration curve look like?
system is supplied by a
200 kW power plant having • The definition of the equivalent load is:
an availability of 80% and
the operation cost 1
¤/kWh. Calculate the EENS,
LOLP and ETOC of the
system.
• This means convolution since the equivalent load
is a sum of independent stochastic variables
19 20
Calculation of equivalent load
Calculation of system indices
load duration curve, ELDC
• General formula: • Example 6.4-6.5. Study the same system as in
example 6.3. What does the equivalent load
duration curve look like?

• Assume an LDC and one power plant with

availability p1  ELDC:

• The equivalent load is > x, when

- unit available, load is > x, probability p1
- outage of G1, load > x-G1, probability q1

21 22

Calculation of system indices Calculation of system indices

• Example 6.6. Calculate the unserved
energy, loss of load probability and
expected operation cost per hour in the
system described in figure 6.3.

23 24
 Example 6.6
Example 6.6

25 26

Example 6.6

27
 Power Generation Operation PPC model
and Planning, EG2200
Assume
• Perfect competition
Probabilistic production cost • Perfect information
simulation of electricity • Load is not price sensitive
markets – lecture 14 • Neglect grid losses and
limitations
Lennart Söder
• All scenario parameters can be
Professor in Electric Power Systems
treated as independent

Some of these assumption can be

treated with some specific methods.

1 2

Calculation of system index Calculation of system index

- Basic idea - Basic idea
LOLP = LDC(600) = 876 h/year = 10%

h/year LDC h/year LDC

3 4
Calculation of system index
Load model – Alternative 2
- Basic idea
Example 6.8:

Perfect competition  Unit 1 is used first, since ß1  ß2

h/year LDC

The area is now changed since the y-axes is divided with T. To get the
correct expected value, the area has to be multiplied with T.

5 6

Thermal power station model Equivalent load

Definition 6.10. The equivalent load is
given by

where
• The availability of a thermal power station
Eg = equivalent load for the power plant
can not be calculated, but has to be next to be dispatched after unit g,
estimated from historical data and D = actual load,
forecasts of the future. Ok = outage in unit k.

7 8
Calculation of system indeces
• Risk of capacity deficit = Loss of load
probability. Capacity deficit occurs when
the deman exceeds available production
capacity, i.e.
9 10
Calculation of system indeces
• Expected energy production in power
plants, EG: Assume that we do not have
unit g. Then the expected unserved energy
is . When unit g is then installed
the expected unserved energy decreases to
The difference is expected energy
production in the unit.
11
Calculation of system indeces
• Expected Energy Not Served, EENS: This
means energy that cannot be delivered
depending on capacity deficit, i.e.,
equivalent load > installed capacity
Calculation of system indeces
• Example 6.4. Study the same system as in
example 6.3. What does the equivalent load
duration curve look like?
• The definition of the equivalent load is:
• This means convolution since the equivalent load
is a sum of independent stochastic variables
12
Calculation of equivalent load Model of a wind power station
load duration curve, ELDC The production varies when the wind varies
• General formula:

• Assume an LDC and one power plant with

availability p1  ELDC:

• The equivalent load is > x, when

- unit available, load is > x, probability p1
- outage of G1, load > x-G1, probability q1

13 14

Model of a wind power station

Model of a wind turbine
The wind varies
• There is one discrete probability that the
wind power plant will not produce
anything, i.e. When there is too little wind,
too much wind or there is an outage.
• There is one discrete probability that the
wind power plant will produce installed
capacity, i.e. When the wind is higher than
rated wind speed and lower than cut-off
wind speed, and there is no outage.
• During the rest of the time there is a
continous distribution between 0 MW and
installed capacity

15 16

Model of a wind turbine
Wind power probability function
fp(x)
Model of many wind turbines
• The production in several wind power plants
can not be treated as independent variables,
Thermal power probability function since high wind in one place normally means
high winds in neighbouring regions

0
0 C
[MW]

17 18

Model of many wind turbines Model of wind power

• Since PPC is based on independence • Available wind power capacity is a
between unit variation (outages) in continous stochastic variable  use
different power plants, all wind power has discrete approximation.
to be treated as one source:
• Compared to one unit pdf, the discrete real
probabilities are much smaller
of
fp(x)

0
0 C
[MW] 19 20
Model of wind power PPC model
• Thermal power is treated with two states:
Installed capacity or zero. Assume
• Total wind power is treated with a multi- • Perfect competition
state unit: • Perfect information
• ELDC Convolution with thermal power: • Load is not price sensitive
• Neglect grid losses and
limitations
• All scenario parameters can be
treated as independent
• Wind power:
Some of these assumption can be
treated with some specific methods.

21 22

Wind power in PPC

• ”All scenario parameters can be treated as
independent”
• Outages in thermal power plant can be
considered to be independent on wind
speed
• There is often a common variation between
load and wind:
time coupling: treat as independent in
each period
special couplings (e.g. low wind when it
is extremely cold): treat as independent in
each period
Wind power in PPC
• Example 6.13 (the power system in a
small island): The small island Kobben is
not connected to the national grid, but
there is a local grid which is powered by a
small wind power plant (installed capacity
200 kW) and a diesel generator set. The
diesel generator set has a maximal
capacity of 200 kW and the availability is
95%. A simplified model of the wind power
plant is stated in table 6.5. The load
duration curve is shown in figure 6.12.
Calculate the risk of power deficit in this
system.

23 24
Example 6.13 Example 6.13

25 26

Power market simulations

- Capacity credit of a plant Wind Power Capacity Credit
(expressed as equivalent load increase)

• Definition: Capacity credit means

How much can the consumption
the possibility of a power plant to
increase when the amount of wind
increase the reliability (decrease in
power increases and the risk of
LOLP) of a plant
power deficit is kept constant?
• Question: Is there any capacity
credit for wind power?

27 28
Wind power capacity credit - 1 Wind power capacity credit - 2

No wind power With wind power

29 30

Wind power capacity credit - 3 Capacity credit of wind power

• ”True” value: Considers the

possibility of wind power increase the
reliability of the power system, 
20% of installed capacity.
• Market value: High market prices
when there is a risk for power deficit.

With wind power, load + 300 MW

31 32