Unit Commitment

Unit Commitment (UC)

• UC determines the ON/OFF status of generators for
different intervals over a pre-specified time period T
to meet the forecasted loads and other constraints
such that the total cost is minimized.
• T can be
• as short as 1-2 hours (e.g., intermediate term dispatching)
• as long as years (e.g., long-term planning)
• Typically, T is 24 hours covering the next operating day,
used for day-ahead scheduling and day-ahead markets.
--- specifically, this is called day-ahead unit commitment.
• With system reliability constraints, it is also called
security-constrained UC (SCUC).
2
 Unit Commitment
• Why SCUC?
• Long startup time
• Minimum up and down time
• Pmin > 0
• One-time cost (e.g., startup cost, shutdown cost)
• Other fixed costs (e.g., no-load cost)
• SCUC: Binary variables are required to capture
these characteristics by indicating the on/off
status of each unit for each interval.

3
 Unit Commitment
• Inputs
• Load profiles for the next operating day
• Available units
• network information and system requirements
• Model
• Mixed-integer linear programming (MILP) problem.
• Outputs/results
• which intervals (hours) should a unit be ON
(connected to the system)
• how much power should each unit produce during
the in-service hours.
4
 Unit Commitment
• If way too few units committed
• Infeasibility: failure to meet the load
• If way too many units committed
• Infeasibility: when the sum of Pg,min for all
committed units is larger than total demand.
• For feasible unit commitment solutions:
• Non-optimal combinations: unnecessary extra cost.
• Optimal combinations: minimum cost while meeting
demands and all constraints.

5
• Example 1
Unit Commitment
• Generator 1: , ,
• Generator 2: , ,
• Total load: 70 MW. Two units initially are off.
• Infinite network capacity.
• No minimum up time and down time requirements
• No no-load cost.
• One single time interval.
Unit 1 Unit 2
How to solve this example? Combination 1 OFF OFF
• Let’s try enumeration Combination 2 OFF ON
Combination 3 ON OFF
For each possible
combination, we run DCOPF. Combination 4 ON ON

6
• Example 1
Unit Commitment
• Generator 1: , ,
• Generator 2: , ,
• Total load: 70 MW.

Pg1 Pg2 Total cost

Unit 1 Unit 2 Feasible
(MW) (MW) ($)
Soln 1 OFF OFF No
Soln 2 OFF ON Yes 0 70 70*30 + 100 = 2,200
Soln 3 ON OFF Yes 70 0 70*10 + 800 = 1,500
(50*10 + 800) + (20*30 +
Soln 4 ON ON Yes 50 20
100) = 2,000

7
• Example 2
Unit Commitment
• Generator 1: , ,
• Generator 2: , ,
• Generator 3: , ,
• Total load: 150 MW. Three units initially are off.
• Infinite network capacity.
• No minimum up time and down time requirements
• No no-load cost.
• One single time interval.

Enumeration: combinations

8
• Example 2
Unit Commitment
• Generator 1: , ,
• Generator 2: , ,
• Generator 3: , ,
• Total load: 150 MW
Unit 1 Unit 2 Unit 3 Feasible Pg1 Pg2 Pg3 Total
(MW) (MW) (MW) cost ($)
Combination 1 OFF OFF OFF No
Combination 2 OFF OFF ON No
Combination 3 OFF ON OFF No
Combination 4 OFF ON ON YES 0 90 60 4,600
Combination 5 ON OFF OFF No
Combination 6 ON OFF ON No
Combination 7 ON ON OFF Yes 80 70 0 3,800
Combination 8 ON ON ON Yes 80 20 50 3,850
9
• Example 3
Unit Commitment
• Generator 1: , ,
• Generator 2: , ,
• Two time intervals.
• Total load: 70 MW for hour 1; 110 MW for hour 2.
• Infinite network capacity. Two units initially are off.
• No minimum up time and down time requirements.
• No no-load cost.

For enumeration method, number of possible combinations

(including infeasible combinations)?
𝑁𝐺 2
2 ∙ 𝑁 𝑇 =2 ∙ 2= 8
𝑁 𝐺 𝑁𝑇 2 2 2
(2 ) =(2 ) = 4 =16
10
• Example 3
Time interval 1 Time interval 2
Unit 1 Unit 2 Unit 1 Unit 2
Combination 1 OFF OFF OFF OFF
Combination 2 OFF ON
Combination 3 ON OFF
Combination 4 ON ON • There are 4 different
Combination 5 OFF ON
combinations for time
OFF OFF
interval 1;
Combination 6 OFF ON
• for each of them, there are 4
Combination 7 ON OFF combinatorial solutions for
Combination 8 ON ON time interval 2.
Combination 9 ON OFF OFF OFF • In total, 16 different possible
Combination 10 OFF ON combinations.
Combination 11 ON OFF
Combination 12 ON ON
Combination 13 ON ON OFF OFF
Combination 14 OFF ON
Combination 15 ON OFF
Combination 16 ON ON 11
 Enumeration for SCUC
• For each unit in each interval: two possible status.
• Examine all possible combinations of unit status
• Is it practical?
• Number of generators for practical power systems
• ~350 units for the TVA system
• ~700 units for the ERCOT system
• ~2,800 units for the PJM system

12
 Enumeration for SCUC
• TVA unit commitment problem
• ~350 units, 1 hour
• # of binaries: 350*1=350
• # of combinations:
• To solve SCUC with enumeration method:
• Assume: DCOPF can solve 1 million cases per second.
• It will take seconds or years.
• Impossible!
• In reality, day ahead SCUC covers 24 hours.
• The time required to solve SCUC is only 1-4 hours.
• Thus, enumeration method is NOT practical.
13
 SCUC Methods
• Dynamic Programming
• can provide a feasible solution very quickly.
• subject to curse of dimensionality.
• Lagrange Relaxation (LR)
• provide an optimality gap to gauge solution quality.
• may not provide a feasible solution in some cases.
• (was) widely used in ISOs.
• many variations/enhanced versions of original LR.
• Branch and Bound – Mixed Integer Programming (MIP)
• essentially, a method of “divide-and-conquer”.
• provide an optimality gap.
• may quickly identify a feasible solution.
• may provide a globally optimal solution.
14
 Optimality Gap
• Defined as difference between the objective value of a
best known feasible solution and a value that bounds the
best possible objective value.
• Why optimality gap?
• Given an optimization problem, it may not be solved to the
actual optimal solution, which is true for most large-scale
practical problems.
• Instead, we may only obtain a “good” feasible solution.
Optimality gap will measure how good this “good” solution is.
• Note that most people call this “good” solution as “optimal
solution” though it is not the actual optimal solution.
• Two gaps: absolute gap and relative gap.
15
 Optimality Gap
• For minimization problems
• Upper bound (UB): objective value of a best known feasible
solution.
• Lower bound (LB): a possible minimal objective value.
• Absolute gap: (UB – LB)
• Relative gap: (UB – LB)/LB
• Global optimal solution: gap is 0.
Objective values
Solver always tries to find a better feasible
solution with a lower objective value (UB).
UB Best known feasible solution
The optimal objective value can be anywhere in this range [LB, UB].
LB
Dual problem will provide a bound for the minimal
16
value and it always looks for a higher bound (LB).
 Industrial Practice
• Typical setting by ISOs
• Relatively MIP gap: ~0.1%
• Absolute MIP gap: ~$5,000
• SCUC stops when
• a solution met both gaps is found,
• Or, it reaches the pre-specified maximum iterations/time.
• Algorithms
• Previously, Lagrange relaxation (LR) was used by ISOs.
• Now mainly use Mixed Integer Programming (MIP).
• LR can serve as a backup method in case MIP does not solve.
• Popular commercial MIP solvers: Gurobi and Cplex.

17
SCUC Model
 SCUC Model
Notation

19
 Objective function
SCUC Model
𝑚𝑖𝑛❑ ∑ ∑ (𝑐 𝑔 𝑃 𝑔𝑡 +𝑐 𝑔 𝑢 𝑔𝑡 +𝑐 𝑔 𝑣 𝑔𝑡 )
𝑁𝐿 𝑆𝑈

𝑔∈ 𝐺 𝑡 ∈𝑇
Constraints

∑ 𝑃 𝑔𝑡 + ∑ 𝑃 𝑘𝑡 − ∑ ¿
∈ 𝐺(𝑛) 𝑘∈ 𝐾 (𝑛 −) 𝑘 ∈ 𝐾 ¿¿

h𝑟
𝑃 𝑔𝑡 − 𝑃 𝑔,𝑡 −1 ≤ 𝑅 ∀ 𝑔 , 𝑡
𝑔
h𝑟
𝑃 𝑔, 𝑡 − 1 − 𝑃 𝑔𝑡 ≤ 𝑅 ∀ 𝑔 , 𝑡
𝑔
20
 SCUC Objective Function
Objective function
𝑚𝑖𝑛❑ ∑ ∑ (𝑐 𝑔 𝑃 𝑔𝑡 +𝑐 𝑁𝐿 𝑆𝑈
𝑔 𝑢 𝑔𝑡 +𝑐 𝑔 𝑣 𝑔𝑡 )
𝑔∈ 𝐺 𝑡 ∈𝑇
is to minimize the total cost including (i) operating cost,
(ii) no load cost, (iii) start-up cost.
Parameters: Variables:
: production cost ($/MWh) : output of unit at interval .

: no-load cost ($) : status of unit at interval .

: Startup indicator of unit at
: one-time startup cost ($)
interval .

and are binary variable while is continuous variable.

21
 SCUC Constraints
Generator constraints:
Unit status indicator: 1 means
online; 0 means offline.
Minimum limits for online
units.
Maximum limits for online units.
Generator startup indicator: 1 means
unit is turned on at the associated time
interval.
Generator startup variables are
constrained by the commitment
variables
𝑃 𝑔𝑡 − 𝑃 𝑔,𝑡 −1 ≤ 𝑅𝑔 ∀ 𝑔 , 𝑡
Generator ramping rate limit constraints
𝑃 𝑔, 𝑡 − 1 − 𝑃 𝑔𝑡 ≤ 𝑅𝑔 ∀ 𝑔 , 𝑡
22
 SCUC Model
Example to illustrate the relationship between and

Hours that unit are on

Hour t 1 2 3 4 5 6 7 8 9
ON/OFF status 0 0 1 1 1 1 1 0 0
0 0 1 0 0 0 0 0 0
…
Unit startup
Unit shutdown 0 0 0 0 0 0 0 1 0

Unit shutdown indicator

Some models use while others do not.

23
 SCUC Model
Network constraints:
Branch thermal limits

Branch power flow calculation

∑ 𝑃 𝑔𝑡 + ∑ 𝑃 𝑘𝑡 − ∑ ¿ Nodal power balance equations

𝑔 ∈ 𝐺(𝑛) 𝑘∈ 𝐾 (𝑛 −) 𝑘 ∈ 𝐾 ¿¿

24
• Example 4
Unit Commitment
• Generator 1: , , ,
,
• Generator 2: , , ,
,
• Two time intervals.
• Total load: 70 MW for hour 1; 110 MW for hour 2.
• Infinite network capacity and no losses.
• No minimum up time and down time requirements.
• Two generators are off initially.
• No start-up ramping limit and no shut-down ramping limit.

25
• Example 4

26
• Example 4
Unit Commitment
• Generator 1: , , ,
,
• Generator 2: , , ,
,
• Load of two time intervals: 70 MW for hour 1; 110 MW for hour 2.
AMPL Results:

27
• Example 4
Unit Commitment
• Generator 1: , , ,
,
• Generator 2: , , ,
,
• Load of two time intervals: 70 MW for hour 1; 110 MW for hour 2.

Model-2:
Another form of SCUC program in AMPL.

28