Download as pdf or txt
Download as pdf or txt
You are on page 1of 18

ECE 6379

Power System Operations and


Modeling
College of Engineering
University of Houston

Professor Xingpeng Li
Unit Commitment Solution Methods

2
Advanced SCUC Methods
• Dynamic Programming (DP)
• can provide a feasible solution very quickly.
• subject to curse of dimensionality.
• Lagrange Relaxation (LR)
• provide an optimality gap to gauge solution quality.
• (was) widely used in ISOs.
• may not provide a feasible solution in some cases.
• many variations/enhanced versions of original LR.
• Branch and Bound (BB) – 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. 3
Optimality Gap
• Defined as the 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.
4
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 could be anywhere in this range [LB, UB].
LB
Dual problem will provide a bound for the minimal
5
value and it always looks for a higher bound (LB).
Industrial Practice
• 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.
• Previously, LR method was used by ISOs.
• Today, ISOs mainly or only use MIP method; some
ISOs may use LR as a backup plan.
• Solvers: Cplex or Gurobi.

6
Dynamic Programming

7
Dynamic Programming
A prototype example for dynamic programming:

Find the lowest total path cost from A to J


Is path A→B → F → I → J (for a total cost of 13) the cheapest route?
No! There are cheaper options.
8
Ref: Frederick S. Hillier, Gerald J. Lieberman, “Introduction to operations research”, [Book], 10th Edition, ISBN 978-0-07-352345-3, MHID 0-07-352345-3
Dynamic Programming

• Trial-and-error solution
• Very time consuming for large problems.
• Dynamic programming (DP) solution
• Starts with a small portion of original problem.
• Finds optimal solution for this smaller problem.
• Gradually enlarges the problem.
• Finds the current optimal solution from the preceding one.

9
Ref: Frederick S. Hillier, Gerald J. Lieberman, “Introduction to operations research”, [Book], 10th Edition, ISBN 978-0-07-352345-3, MHID 0-07-352345-3
Dynamic Programming
Find the lowest total path cost from A to J

Problem formulation
• Decision variables 𝑥1 , 𝑥2 , 𝑥3 , 𝑥4
• Route begins at A, proceeds through 𝑥1 , 𝑥2 , 𝑥3 , 𝑥4 , and ends at J.
Procedure:
• Forward DP algorithm.
• Backward DP algorithm.
10
Ref: Frederick S. Hillier, Gerald J. Lieberman, “Introduction to operations research”, [Book], 10th Edition, ISBN 978-0-07-352345-3, MHID 0-07-352345-3
Dynamic Programming

The state at the next stage is completely determined by


the current stage and the policy decision at that stage.

11
Ref: Frederick S. Hillier, Gerald J. Lieberman, “Introduction to operations research”, [Book], 10th Edition, ISBN 978-0-07-352345-3, MHID 0-07-352345-3
Dynamic Programming
Forward DP algorithm

S-1 S-2 S-3 S-4 S-5

What is the least cost from A to B, C, and D in Stage-1 respectively?

12
Dynamic Programming
Forward DP algorithm

See notes for the solution.

13
Dynamic Programming
Backward DP algorithm

S-1 S-2 S-3 S-4 S-5

What is the least cost from J to H and I in Stage-4 respectively?

14
Dynamic Programming
Backward DP algorithm

See notes for the solution.

15
Dynamic Programming
Features of Dynamic Programming:
• Problem can be divided into stages with a policy
decision required at each stage.
• Each stage has a number of states.
• The policy decision at each stage transforms the
current state into a state associated with the beginning
of the next stage.
• Solution procedure designed to find an optimal policy
for the overall problem.

16
Ref: Frederick S. Hillier, Gerald J. Lieberman, “Introduction to operations research”, [Book], 10th Edition, ISBN 978-0-07-352345-3, MHID 0-07-352345-3
Dynamic Programming
SCUC Example:
• Unit 1: 30 ≤ 𝑃1 ≤ 80 𝑀𝑊, 𝐶1 = 10 $/𝑀𝑊ℎ, 𝑆𝑈1 = $800
• Unit 2: 50 ≤ 𝑃2 ≤ 90 𝑀𝑊, 𝐶2 = 30 $/𝑀𝑊ℎ, 𝑆𝑈2 = $100
• Total load: 90 MW in the first hour and 70 MW for the second hour.
Using forward dynamic programming to solve this two-interval unit
commitment problem? The objective is to minimize the total cost.
Both units are initially offline.

17
Dynamic Programming
SCUC Example:

See notes for the solution.

18

You might also like