Unit Commitment

Economic Dispatch: Problem Definition

 Given load
 Given set of units on-line
 How much should each unit generate to meet this load at
minimum cost?

L
A B C

2
Typical summer and winter loads

Unit Commitment
 Given load profile
(e.g. values of the load for each hour of a day)
 Given set of units available
 When should each unit be started, stopped and how much
should it generate to meet the load at minimum cost?

? ? ?
Load Profile
G G G

4
Data for example
Unit1 Min  150MW
Max  600 MW
H  5100  72P  000142P2 MBtuh
1 1 1

Unit2  Min  100MW

Max  400 MW
H  3100
2  785P2  000194P22 MBtuh
Unit3 Min  50 MW
Max  200 MW
H  780  797P  000482P2 MBtuh
3 3 3

with fuel costs:

Fuel cost1  11RMBtu
Fuel cost2  10RMBtu
Fuel cost3 12RMBtu
What combination of on line units should be used to supply 550 MW?

Unit Combinations to supply 550 MW

6
 Simple peak-valley load pattern

“Shut-down Rule”
Optimum Combination
Load Unit 1 Unit 2 Unit 3
1200 On On On
1150 On On On
1100 On On On
1050 On On On
1000 On On Off
950 On On Off
900 On On Off
850 On On Off
800 On On Off
750 On On Off
700 On On Off
650 On On Off
600 On Off Off
550 On Off Off
500 On Off Off

When load is above 1000 MW, run all three units; between 1000 MW  
and 600 MW, run units 1 and 2; below 600 MW, run only unit 1.

8
Shut down Rule applied to load pattern

Observations on the example:

 Far too few units committed:
Can’t meet the demand

 Not enough units committed:

Some units operate above optimum

 Too many units committed:

Some units below optimum

 Far too many units committed:

Minimum generation exceeds demand

 No-load cost affects choice of optimal combination

10
A more ambitious example
 Optimal generation schedule for
a load profile
Load
 Decompose the profile into a
set of periods 1000

 Assume load is constant over 500

each period
Time
 For each time period, which 0 6 12 18 24
units should be committed to
generate at minimum cost
during that period
11

Optimal combination for each hour

12
Matching the combinations to the load

Load
Unit 3

Unit 2

Unit 1

Time
0 6 12 18 24

13

Issues
 Must consider constraints
 Unit constraints
 System constraints
 Some constraints create a link between periods
 Start-up costs
 Cost incurred when we start a generating unit
 Different units have different start-up costs
 Curse of dimensionality
 Crew Constraints
 Fuel Constraints

14
Unit Constraints
 Constraints that affect each unit individually:
 Maximum generating capacity
 Minimum stable generation
 Minimum “up time”
 Minimum “down time”
 Ramp rate

15

Notations

u(i,t) : Status of unit i at period t

u(i,t)  1 : Unit i is on during period t

u(i,t)  0 : Unit i is off during period t

x(i,t) : Power produced by unit i during period t

16
Minimum up- and down-time
 Minimum up time
 Once a unit is running it may not be shut down immediately:

 Minimum down time

 Once a unit is shut down, it may not be started immediately
If u(i,t)  1 and t iup  tiup,min then u(i,t  1)  1

If u(i,t)  0 and tidown  tidown,min then u(i,t  1)  0

17

Ramp rates
 Maximum ramp rates
 To avoid damaging the turbine, the electrical output of a unit
cannot change by more than a certain amount over a period of
time:

Maximum ramp up rate constraint:

x  i,t  1  x  i,t   Pi up,max

Maximum ramp down rate constraint:

x(i,t)  x(i,t  1)  Pi down,max

18
System Constraints
 Constraints that affect more than one unit
 Load/generation balance
 Reserve generation capacity
 Emission constraints
 Network constraints

 Spinning reserve: enough generation  operated below generator 
limit to allow  generators to rapidly make up loss of one or  more
generators

19

How much reserve?

 Protect the system against “credible outages”
 Deterministic criteria:
 Capacity of largest unit or interconnection
 Percentage of peak load
 Probabilistic criteria:
 Takes into account the number and size of the committed units as
well as their outage rate

20
Types of Reserve
 Spinning reserve
 Primary
 Quick response for a short time
 Secondary
 Slower response for a longer time
 Tertiary reserve
 Replace primary and secondary reserve to protect against another
outage
 Provided by units that can start quickly (e.g. open cycle gas turbines)
 Also called scheduled or off-line reserve

21

Types of Reserve
 Positive reserve
 Increase output when generation < load
 Negative reserve
 Decrease output when generation > load

 Other sources of reserve:

 Pumped hydro plants
 Demand reduction (e.g. voluntary load shedding)

 Reserve must be spread around the network

 Must be able to deploy reserve even if the network is congested

22
Cost of Reserve
 Reserve has a cost even when it is not called
 More units scheduled than required
 Units not operated at their maximum efficiency
 Extra start up costs
 Must build units capable of rapid response
 Cost of reserve proportionally larger in small systems
 Important driver for the creation of interconnections between systems

23

Example 4C Spinning Reserve Allocation

 Maximum Transfer Capacity of 550 MW
 Loss of a Unit in Western Region ?
 Loss of a Unit in Eastern Region ?

24
Environmental constraints
 Scheduling of generating units may be affected by
environmental constraints
 Constraints on pollutants such SO2, NOx
 Various forms:
 Limit on each plant at each hour
 Limit on plant over a year
 Limit on a group of plants over a year
 Constraints on hydro generation
 Protection of wildlife
 Navigation, recreation

25

How Many Combinations Are There?

111  Examples
110  3 units: 8 possible states
101  N units: 2N possible states
100

011

010

001

000

26
How Many Solutions Are There?
 Optimization over a time horizon
divided into intervals

 Assuming a single unit can also

supply the total load

 Combinations to try in each hour

,1 ,2 .. , 2 1

T= 1 2 3 4 5 6

27

How many solutions are there anyway?

 Optimization over a time

horizon divided into intervals

 A solution is a path linking one

combination at each interval

 How many such path are

there?
2 1
T= 1 2 3 4 5 6

28
 The Curse of Dimensionality
 We must establish a loading pattern for M   periods.
 We have  N  units to commit and dispatch.
 The M load levels and operating limits on the N units are such 
that any one  unit can supply the individual loads and that any 
combination of units can also  supply the loads.

where   , is the combination of  N items taken j  at 
a time. That is,
!
,
! !

29

The Curse of Dimensionality

 Example: 5 units, 24 hours

 2 1 2 1 6.204 10

 Processing 109 combinations/second, this would take

1.967 × 1019 years to solve

 There are 100’s of units in large power systems...

 Many of these combinations do not satisfy the constraints

30
How do you Beat the Curse?
Brute force approach won’t work!

 Need to be smart
 Try only a small subset of all combinations
 Can’t guarantee optimality of the solution
 Try to get as close as possible within a reasonable amount of
time

31

Main Solution Techniques

 Characteristics of a good technique
 Solution close to the optimum
 Reasonable computing time
 Ability to model constraints

 Priority list / heuristic approach

 Dynamic programming
 Lagrangian relaxation
 Mixed Integer Programming
State of the art

32
 Priority List Solution
Full Load
Unit Average Production Cost ( R /MWh)
1 9.79
2 9.48
3 11.188

Unit R /MWh Min MW Max MW

Priority order
2 9.48 100 400
1 9.79 150 600

3 11.188 Min MW50from Max200

MW
Combination Combination Combination
2 1 3 300 1200
2 1 250 1000
2 100 400

3
3

Dynamic Programming Paths and

Unit Commitment
Each state represents a combination of generating  
units supplying the load specified for that hour. Each  
state has a production cost PCOST

PCOST
STATES

PCOST FCOST is the accumulated cost  
to get to a state from the start  
scost PCOST through optimum path leading  
scost to that state
A dot represents one state  
operating in one time period

TIME PERIODS

FCOST(end of path) =
PCOST(start of path) + SCOST(along path) +FCOST(start of path)

3
4
Unit Data

Min Min No-load Marginal Start-up

Pmin Pmax Initial
Unit up down cost cost cost
(MW) (MW) status
(h) (h) ($) ($/MWh) ($)

A 150 250 3 3 0 10 1,000 ON

B 50 100 2 1 0 12 600 OFF

C 10 50 1 1 0 20 100 OFF

35

Demand Data

Hourly Demand
350
300
250
200
Load
150
100
50
0
1 2 3
Hours

Reserve requirements are not considered

36
Feasible Unit Combinations (states)
Combinations 1 2 3
Pmin Pmax
A B C 150 300 200
1 1 1 210 400
1 1 0 200 350
1 0 1 160 300
1 0 0 150 250
0 1 1 60 150
0 1 0 50 100
0 0 1 10 50
0 0 0 0 0
37

Transitions between feasible combinations

1 2 3
A B C
1 1 1
1 1 0
1 0 1
1 0 0 Initial State

0 1 1

38
Infeasible transitions: Minimum down time of
unit A
1 2 3
A B C
1 1 1
1 1 0
1 0 1
1 0 0 Initial State

0 1 1
TD TU
A 3 3
B 1 2

39
C 1 1

Infeasible transitions: Minimum up time of

unit B
1 2 3
A B C
1 1 1
1 1 0
1 0 1
1 0 0 Initial State

0 1 1
TD TU
A 3 3
B 1 2

40
C 1 1
Feasible transitions
1 2 3
A B C
1 1 1
1 1 0
1 0 1
1 0 0 Initial State

0 1 1

41

Operating costs

1 1 1 4

1 1 0 3 7

1 0 1
2 6
1 0 0 1
5

42
Economic dispatch
State Load PA PB PC Cost
1 150 150 0 0 1500
2 300 250 0 50 3500
3 300 250 50 0 3100
4 300 240 50 10 3200
5 200 200 0 0 2000
6 200 190 0 10 2100
7 200 150 50 0 2100
Unit Pmin Pmax No-load cost Marginal cost
A 150 250 0 10
B 50 100 0 12
43 C 10 50 0 20

Operating costs

1 1 1 4
$3200

1 1 0 3 7
$3100 $2100

1 0 1 2 6
$3500 $2100

1 0 0 1 5
$1500 $2000

44
Start-up costs

1 1 1 4
$3200
$0
1 1 0 $700 3 $0 7
$3100 $600 $2100
$600
1 0 1 2 $0 6
$100 $3500 $2100
$0
1 0 0 $0 1 5
$1500 $2000
Unit Start-up cost

A 1000
B 600
45 C 100

Accumulated costs
$5400
1 1 1 4
$3200
$0
$5200 $7300
1 1 0 $700 3 $0 7
$3100 $600 $2100
$600 $5100 $7200
1 0 1 2 $0 6
$3500 $2100
$100
$0
$1500 $7100
1 0 0 $0 1 5
$1500 $2000

46
Total costs

1 1 1 4

$7300
1 1 0 3 7

$7200
1 0 1 2 6

$7100
1 0 0 1 5

Lowest total cost

47

Optimal solution

1 1 1

1 1 0

1 0 1 2

$7100
1 0 0 1 5

48
Dynamic Programming
Solution

Schedule shown is: 111  interval 1, 110 interval 2, 100 interval 3, 101 interval 4

49

Path Multiplication

50
Control paths enumeration using reduced
search range

51

Recursive Algorithm
 The following assumption are made
 A state consists of an array of units with specified units operating
and the rest off-line.
 The start-up cost of a unit is independent of the time it has been off-
line.
 There are no costs for shutting down a unit.
 There is a strict priority order, and in each interval a specified
minimum amount of capacity must be operating.

 A feasible state is on in which the committed units can supply

the required load and that meets the minimum amount of
capacity each period
52
Forward Dynamic Programming
 A recursive algorithm to compute the minimum cost in hour K
with combination I

 State (K, I) is the Ith combination in hour K

53

X is the number of states to search each period

N is the number of strategies to save at each step

54
Dynamic Programming
Example

Simplified generator cost

function
Single Step and Multiple Step Incremental Cost
Curve

57

Full set of unit combinations

a, 1=on
0=off

20
Case 1, Strict priority order Dispatch:

 Total Cost for the interval is the sum of eight dispatches plus
the transitional cost for starting any units.
 Only the following states are examined.

 All possible commitments start from state 12 since this was

given as the initial condition.

59

Sample Calculations – Case 1

60
Sample Calculations – Case 1

61

Case 2 – Complete Enumeration

 A limit of 2 1 15 dispatches for each of the 8 h
 Theoretical maximum of 15 2.56 10 possibilities.
 Most of these are not feasible because they do not supply
sufficient capacity and can be discarded with little analysis
required.

62
63

Results

6
© Bruce F. Wollenberg, University of Minnesota
4
Case 3: Using Minimum Shut Down and
Operating Times
 The upper bound on X shown in the flowchart was 15.
 Three different values for N, the number of strategies saved at
each stage, were taken as 4, 8, and 10.
 The same trajectory was found for values of 8 and 10.
 When only four strategies were saved, the procedure
flounders (i.e., fails to find a feasible path) in hour 8, because
the lowest cost strategies in hour 7 have shut down units that
cannot be restarted in hour 8 because of minimum unit
downtime rules.

65

Case 3: Using minimum shut down rules

6
6
Summary

67

Flexible Plants
 Power output can be adjusted (within limits)
 Examples:
 Coal-fired
 Oil-fired
 Open cycle gas turbines
 Combined cycle gas turbines Thermal units
 Hydro plants with storage

 Status and power output can be optimized

68
Inflexible Plants
 Power output cannot be adjusted for technical or commercial
reasons
 Examples:
 Nuclear
 Run-of-the-river hydro
 Renewables (wind, solar,…)
 Combined heat and power (CHP, cogeneration)
 Output treated as given when optimizing

69

Notes
 This example is intended to illustrate the principles of unit
commitment
 Some constraints have been ignored and others artificially
tightened to simplify the problem and make it solvable by hand
 Therefore it does not illustrate the true complexity of the
problem
 The solution method used in this example is based on dynamic
programming. This technique only works well for small systems
(< 20 units)

70
Lagrange Relaxation Solution
 Disadvantages of Dynamic Programming for large systems
 Search over a small number of commitment states

 Some of the disadvantage are addressed in Lagrange Relaxation

Technique

 Define variable Uit

71

Unit Commitment Formulation

 Objective Function

 Constraints
 Loading Constraint

 Unit Limits

 Unit Minimum Up Time and Down Time constraints

72
 Lagrangian Function to minimize

 Coupling Constraints
 One unit affects the other unit
 Non Coupling Constraints
 What happens to one unit does not effect the cost of running the
other unit
 Lagrangian Relaxation Technique solves the UC problem by
temporarily relaxing or ignoring the coupling constraints.

73