Power System Operation and Planning

Lecture - Unit Commitment

Abdul Basit
US Pakistan Centre for Advanced Studies in Energy (USPCAS-E)
University of Engineering & Technology (UET), Peshawar

USPCAS-E 1
Economic dispatch – Summary

• Economic dispatch determines the best way to

minimize the generator operating costs
• Economic dispatch is not concerned with
determining which units to turn on/off (this is
the unit commitment problem)
• Economic dispatch ignores the transmission
system limitations

USPCAS-E 2
Economic dispatch vs Unit commitment

• With a given set of N units running, how of the load much

should be generated at each to cover the load and losses?
This is the question of Economic dispatch.
• The solution is for the current state of the network and does
not typically consider future time periods.
USPCAS-E 3
Economic dispatch vs unit commitment

• Given that there are a number of subsets of N generating

units that would satisfy the expected demand, which of these
subsets should be used in order to provide the minimum
operating cost  Unit Commitment

USPCAS-E 4
Which Generator shall operate?

• Questions? why not just simply “commit” enough units to

cover the maximum system load and leave them running?
• The problem with “commit enough and leave them
running” is a problem of economics. It is quite expensive to
run too many generating units. A lot of money can be saved
by turning units off when they are not needed (when
demand is low). USPCAS-E 5
Hourly consumption – example

USPCAS-E 6
Unit Commitment…
• Some generating units + forecasted load
• Besides the cost of running the units we have additional costs
– start-up cost
– shut-down cost
• Constraints
– spinning reserve
– minimum up and down time
– ramp-up limits, and more
• So: It turns out that we cannot just flip the switch of certain units
on and use them!
– We need to think ahead, and based on the forecasted load and unit
constraints, determine which units to turn on (commit) and which
ones to keep down
– Minimize cost, cheap units play first
– Expensive ones run only when demand is high

USPCAS-E 7
Objective function of the UC
• Objective: Minimize Cost
• Cost is defined using three components
– Fuel Cost
– Start-up Cost
– Shut-down Cost

USPCAS-E 8
UC model - constraints

• Balance between demand and generation

• Generation limits

• Ramp rate constraints

USPCAS-E 9
UC model - constraints

• Minimum Up-Time
– Once the unit is running, it should not be turned off
immediately
• Minimum Down-Time
– Once the unit is de-committed, there is a minimum
time before it can be committed again
• Must-run units
– Some units are given a must-run status due to certain
system requirements (voltage support, steam supply,
etc.)

USPCAS-E 10
UC model - constraints

• Spinning Reserves
– Total generation from synchronized units minus the load
demand and system
– Need spinning reserve to accommodate the load demand
if different from forecast and N-1 contingency events
– How to allocate Spinning reserves?
• Percentage of forecasted peak load demand
• OR to accommodate the loss of heavy loaded units
• OR probability of failure to meet system demand
– Reserves allocation
• Fast generating units and Slow generating units, Activation through AGC

USPCAS-E 11
Spinning Reserves
• Also include off line reserves that are available from quick
starting generators
• Spread in the power system to avoid transmission system
limitation

Example - spinning reserves

USPCAS-E 12
Example – spinning reserves

Need more units to commit in Eastern region to cope

with contingency of 1200 MW unit loss

USPCAS-E 13
UC model - constraints
Thermal unit constraints
• Ramp rate constraints
• Start up time
– Undergoes gradual temperature change and may require some hours to bring unit
online
• Start up cost
– Energy require for changing temperature and pressure lead to Start-Up Cost
• Shut down cost
• Minimum up time
– Once running unit can’t be turned off immediately
• Minimum down time
– Minimum time before the unit is reconnected
• Crew constraint
– Startup require crew members, if there are two units they might not be start at the
same time

USPCAS-E 14
How to solve Unit commitment (UC) problem?

• If a unit is on, we designate this with 1 and respectively, the

off unit is 0
• So, somehow we decide that for the next hour we will have "0
1 1 0 1" if we have five units.
– Based on that, we solve the economic dispatch problem for unit 2, 3 and 5.
• The question now is: how do we come up with this unit
commitment schedule "0 1 1 0 1"?
• One very simplistic way: if we have very few units, go over all
combinations from hour to hour:
– For each combination at a given hour, solve the economic dispatch
– For each hour, pick the combination giving the lowest cost!

USPCAS-E 15
UC Solution Methods

• Complete Enumeration
– gives optimal solution but extremely cumbersome to handle
• Priority List Methods
– the simplest, but may not be optimal
• Dynamic Programming Based Methods
– search technique, can be problematic for large systems
• Langrang Relaxation

USPCAS-E 16
Example A

USPCAS-E 17
 Complete Enumeration method

Case 1 0 0 gives the least cost, However lowers the system reliability
USPCAS-E 18
Priority list method

• Calculate the full load production cost

USPCAS-E 19
Priority list method

• Calculate the full load average production cost

USPCAS-E 20
USPCAS-E 21
Priority List method
Commitment Schedule for 550 MW:
• 1-1-0 (150 MW: 400 MW: 0 MW)
• Total Cost: $5544.55
• Complete Enumeration >> Total cost = $5389

Complete Enumeration Provides the Overall Optimal but May be Extremely

Large to Handle
• For M periods, N units to commit, maximum number of possible
combinations is (2N-1)M. e.g. 10-unit system, 24 hours, this value is
1.73 E 72
• On contrary, priority list method provides only specified combinations
depending on the generating units, e.g. 3 unit systems provides three
combinations.
Priority List is Simple but May be Expensive
USPCAS-E 22
Example B – priority list soln. with start up cost

USPCAS-E 23
Assumptions for Example B

• Unit 2 and Unit 3 are committed initially (at time k = 0)

• The Units have linear cost characteristics

USPCAS-E 24
Load demand forecast

USPCAS-E 25
Priority list solution
• The priority order is decided by the full-load average cost of the units. In
this case, the order is:
Unit-3 >>> Unit-2>>> Unit-1>>>Unit-4
• We ignore minimum up- and down times, and shutdown cost

USPCAS-E 26
UC Solution by Priority List Method, taking into
account shut-down rule
• Step-1: At each hour when load is dropping, determine whether dropping
the next unit on the Priority List will leave sufficient generation to supply
load & spinning reserve?
– No: Continue operating as is
– Yes: Go to Step-2
• Step-2: Determine number of hours H before the unit will be needed
again. That is assuming that the load is dropping and will then go back up
some hours later.
– If H < minimum shut down time for the unit, keep UC as is and go to the last step.
– Else, GO TO Step-3
• Step-3: Calculate:
– Cost C1 = Sum of the hourly production costs for the next H hours with the unit up.
– Cost C2 = The same sum with the unit down and add in the startup cost
– If C2 is less than C1, shut down this unit (OFF), else keep it ON.
• Repeat the above steps for the next unit on Priority List, until no further
units could be shut down. USPCAS-E 27
Priority list method

• Priority list arranged in order of the full load average

cost rate would result in a theoretically correct
dispatch and commitment only if
– No load cost is zero
– Linear input-output characteristics
– Start up costs are a fixed amount

USPCAS-E 28
Dynamic program method

• Assumptions
– A state consists of an array of online units
– For offline units, the start up cost is fixed i.e. independent of
time
– No shut down cost
– Strict priority order in each interval to operate specified amount
of capacity

A feasible state is one in which the committed units can

supply the required load and that meets the minimum
amount of capacity each period

USPCAS-E 29
Forward dynamic program approach
• Developed algorithm run forward in time from the initial
hour to the final hour
– Start up cost is the function of time it has been offline, taking in
account the temperature of the offline unit
– Compares the previous history of unit

The algorithm computes the minimum cost in hour ‘K’ with

combination ‘I’. State (K,I) is the Ith combination in hour K.
The algorithm defines the strategy from one state at a given
hour to a state in the next hour.
The algorithm search for X number of states at each hour
and saves N number of steps at each step

USPCAS-E 30
Lagrang Relaxation solution

• Dynamic programming method reduces the number

of combination by searching over small number of
commitment states >> disadvantage for large power
system
• Langrang Relaxation bases on dual optimization
approach addresses this problem , i.e. Unit
commitment on a power system level
– A variable U is introduced

USPCAS-E 31
Langrang Relaxation - constraints

3. Unit minimum up and down constraints, transmission constraints,

generator fuel limit constraints, system air quality constraints (limits on
emission from fossil fired plants, spinning reserves etc)

USPCAS-E 32
 Langrang equation

Langrang relaxation procedure solve the unit commitment problem

through dual optimization technique, i.e by relaxing or temporary ignoring
the coupling constraints and solving the problem as if they do not exist.
The technique attempts to reach the constrained optimum by maximizing
the ‘λ’ while minimize the langrang equations
USPCAS-E
w.r.t to other variables 33
Initially assume λt as fixed number and later adjust its value

USPCAS-E 34
This separates the units from one another by ignoring the coupling term

The minimum of the Langrangian is found by solving the

minimum of each generating unit over all time periods

USPCAS-E 35
Solution

USPCAS-E 36
Depending on unit limit and optimal point, there are three cases

USPCAS-E 37
Adjusting ‘λ’ in Langrang relaxation technique

• The generating units are scheduled with fixed value

of λ; however, the λ must be adjusted for each hour
to maximize q(λ)

Where

USPCAS-E 38
Adjusting ‘λ’ in Langrang relaxation technique

• The reality duality gap [(J*-q*)/q*] measures the closeness of

solution, where J* is the summation of generation cost over all
periods
• The duality gap when applied to the unit commitment problem, it
finds out

USPCAS-E 39
Langrang relaxation technique

• The duality gap is initially large but progressively becomes

smaller as the iteration progress
• The solution reaches a commitment schedule when at least
enough generation is committed for running economic
dispatch, and further iterations only results in switching
marginal units (ON or OFF)
• The dual solution do not meet the loading constraints
• The final solution is obtained through heuristic algorithm or
restricted dynamic programming to get the final solution (not
limited as in case of dynamic program technique)

USPCAS-E 40
USPCAS-E 41