(EEE-800) Power System Operation, Control and Optimization

EEE-800

Lecture 8: Unit Commitment/Lagrange Relaxation

10:26 AM 1
 Outline
• Unit Commitment Solution Methods
• Dynamic Programming
• Lagrange Relaxation

10:26 AM 2
10:26 AM 3
10:26 AM 4
10:26 AM 5
 Note:
𝑑𝑞 λ
𝛼 = 0.5 𝑓𝑜𝑟 𝑑λ
+ve
𝑑𝑞 λ
𝛼= 0.1 𝑓𝑜𝑟 𝑑λ -ve

using 𝜆 = 2 from previous example ⇒ dual function; 𝑥1 = 4 𝑎𝑛𝑑 𝑥2 = 1 ⇒ J∗ = f x1 , x2 = 5

1st iteration: 2nd iteration:

λ0 =0 𝑑𝑞 λ 5 0
= − (𝜆 ) + 5
5 𝑑λ 2
𝑞 ∗ = 𝑞 λ0 = − (𝜆0 )2 + 5λ0 =0
4
𝑑𝑞 λ
from dual function; 𝑥1 = 0 𝑎𝑛𝑑 𝑥2 = 0 λ1 = λ0 + 𝛼 = 2.5
𝑑λ

10:26 AM 6
 using; λ1 = 2.5; dual function gives; 𝑥1 = 5 𝑎𝑛𝑑 𝑥2 = 1.25

5
𝑞 ∗ = 𝑞 λ1 = − 4 (𝜆1 )2 + 5λ1 = 4.6875

𝑑𝑢𝑎𝑙𝑖𝑡𝑦 𝑔𝑎𝑝 = (J ∗ − 𝑞 ∗ )/q∗ =0.0666

Similarly, repeat this process for next iterations…

10:26 AM 7
9:56 AM 8
10:26 AM 9
10:26 AM 10
Step 2

Is fixed in Step 1

10:26 AM 11
10:26 AM 12
 Ignore the start up cost
-
- -

10:26 AM 13
Step 1

10:26 AM 14
10:26 AM 15
10:26 AM 16
10:26 AM 17
 Example 4E; Solution
𝐹1 𝑃1 = 500 + 10𝑃1 + 0.002𝑃12 ; 100 ≤ 𝑃1 ≤ 600
𝐹2 𝑃2 = 300 + 8𝑃1 + 0.0025𝑃12 ; 100 ≤ 𝑃2 ≤ 400
𝐹2 𝑃3 = 100 + 6𝑃1 + 0.005𝑃12 ; 50 ≤ 𝑃3 ≤ 200

𝑃𝑟𝑖𝑚𝑎𝑙 𝑉𝑎𝑙𝑢𝑒 ⇒ 𝐽∗ = 𝑟𝑒𝑝𝑟𝑒𝑠𝑒𝑛𝑡𝑠 𝑡𝑜𝑡𝑎𝑙 𝑔𝑒𝑛𝑒𝑟𝑎𝑡𝑜𝑟 𝑐𝑜𝑠𝑡 𝑠𝑢𝑚𝑚𝑒𝑑 𝑜𝑣𝑒𝑟 𝑎𝑙𝑙 ℎ𝑜𝑢𝑟𝑠 𝑐𝑎𝑙𝑐𝑢𝑙𝑎𝑡𝑒𝑑 𝑓𝑟𝑜𝑚 𝐸𝐷 𝑝𝑜𝑤𝑒𝑟𝑠.
Note: If not enough generation is committed for any hour then take 𝐽∗ = 10000 𝑓𝑜𝑟 𝑡ℎ𝑎𝑡 ℎ𝑜𝑢𝑟
4
ഥ1 + 𝑈2 + 𝑈3 ))
𝐽∗ = ෍( 500 + 10𝑝1 + 0.002𝑝12 𝑈1 + 300 + 8𝑝2 + 0.0025𝑝22 𝑈2 + 100 + 6𝑝3 + 0.005𝑝32 𝑈3 + 10000 ∗ (𝑈
𝑡=1

1st Iteration

𝜆𝑡 = [0 0 0 0]

𝑈𝑖 𝑡 = 1 min[𝐹𝑖 𝑃𝑖 − 𝜆𝑡 𝑃𝑖𝑡 ]
𝑈𝑖 𝑡 = 0 0

10:26 AM 18
2nd Iteration [Step 1]
𝑇 𝑁
𝑑 𝑑 𝑑
𝑞 𝜆 = L P, U, λ = 𝑡
෍ 𝜆𝑡 (𝑃𝑙𝑜𝑎𝑑 − ෍ 𝑃𝑖𝑡 𝑈𝑖𝑡 )
𝑑𝜆 𝑑𝜆 𝑑𝜆
𝑡=1 𝑖=1

1 2 3 4
𝑡
=σ𝑇𝑡=1 𝜆𝑡 (𝑃𝑙𝑜𝑎𝑑 − σ𝑁 𝑃 𝑡 𝑡
𝑈
𝑖=1 𝑖 𝑖 ) = [P , P , P , P
load load load load ]

= [170,520,1100,330]

⇒ ⇒ [1.7, 5.2, 11, 3.3]

10:26 AM 19
 For; 𝜆 = 5.2 5.2 − 10
2nd
2ndIteration [Step 2]
Iteration 𝑃1 = = −𝑣𝑒 ⇒ 𝑃1𝑚𝑖𝑛 = 100
0.004
5.2 − 8
𝜆𝑡 = [1.7, 5.2, 11, 3.3] 𝑃2 = = −𝑣𝑒 ⇒ 𝑃2𝑚𝑖𝑛 = 100
0.005
𝑑 5.2 − 6
F P = 10 + 0.004P1 = λ 𝑃3 = = −𝑣𝑒 ⇒ 𝑃3𝑚𝑖𝑛 = 50
𝑑𝑃1 1 1 0.01
𝑑
F P = 8 + 0.005P2 = λ
𝑑𝑃2 2 2 For; 𝜆 = 11
11 − 10 𝑜𝑝𝑡
𝑑 𝑃1 = ⇒ 𝑃1 = 250
0.004
F P = 6 + 0.01P3 = λ
𝑑𝑃3 3 3 11 − 8
𝑃2 = = 600 > 𝑙𝑖𝑚𝑖𝑡 ⇒ 𝑃2𝑚𝑎𝑥 = 400
0.005
11 − 6
For; 𝜆 = 1.7 𝑃3 = = 500 > 𝑙𝑖𝑚𝑖𝑡 ⇒ 𝑃3𝑚𝑎𝑥 = 200
0.01
1.7 − 10
𝑃1 = = −2075 ⇒ 𝑃1𝑚𝑖𝑛 = 100
0.004
For; 𝜆 = 3.3 3.3 − 10
1.7 − 8 𝑃1 = = −𝑣𝑒 ⇒ 𝑃1𝑚𝑖𝑛 = 100
𝑃2 = = −1260 ⇒ 𝑃2𝑚𝑖𝑛 = 100 0.004
0.005
3.3 − 8
1.7 − 6 𝑃2 = = −𝑣𝑒 ⇒ 𝑃2𝑚𝑖𝑛 = 100
𝑃3 = = −430 ⇒ 𝑃3𝑚𝑖𝑛 = 50 0.005
0.01
3.3 − 6
𝑃3 = = −𝑣𝑒 ⇒ 𝑃3𝑚𝑖𝑛 = 50
0.01
10:26 AM 20
 𝜆 = 1.7 5.2 11 3.3 𝐹1 𝑃1 − 𝜆1 𝑃1 = 500 + 10𝑃1 + 0.002𝑃12 − 𝜆1 𝑃1
For 𝑃1 = 100 at 𝜆1 =1.7 => 𝐹1 𝑃1 − 𝜆1 𝑃1 =1350
[𝐹1 𝑃1 − 𝜆𝑃1 ] = 1350 1000 375 1190
𝑃1𝑚𝑖𝑛 𝑃1𝑚𝑖𝑛 𝑃1
𝑜𝑝𝑡 𝑃1𝑚𝑖𝑛
𝑈1 = 1 𝜆 = 1.7 5.2 11 3.3
Unit1 𝐹3 𝑃3 − 𝜆𝑃3 = 327.5 152.5 − 700 247.5
𝑃3𝑚𝑖𝑛 𝑃3𝑚𝑖𝑛 𝑃3𝑚𝑎𝑥 𝑃3𝑚𝑖𝑛
𝑈1 = 0
𝑈3 = 1
𝑡=1 𝑡=2 𝑡=3 𝑡=4 Unit3

𝜆 = 1.7 5.2 11 3.3 𝑈3 = 0

𝐹2 𝑃2 − 𝜆𝑃2 = 955 605 − 500 795
𝑡=1 𝑡=2 𝑡=3 𝑡=4
𝑃2𝑚𝑖𝑛 𝑃2𝑚𝑖𝑛 𝑃2𝑚𝑎𝑥 𝑃2𝑚𝑖𝑛
𝑈2 = 1
Unit2

𝑈2 = 0

𝑡=1 𝑡=2 𝑡=3 𝑡=4

10:26 AM 21
You can calculate in Matlab
𝑡
𝑞 𝜆 = 500 + 10𝑝1 + 0.002𝑝12 𝑈1 + 300 + 8𝑝2 + 0.0025𝑝22 𝑈2 + 100 + 6𝑝3 + 0.005𝑝32 𝑈3 + 𝑠𝑢𝑚(𝜆𝑡 (𝑃𝑙𝑜𝑎𝑑 − 𝑠𝑢𝑚(𝑃𝑖𝑡 𝑈𝑖𝑡 ))

3rd Iteration [Step 1]

𝑇 𝑁
𝑑 𝑑 𝑑
𝑞 𝜆 = L P, U, λ = 𝑡
෍ 𝜆𝑡 (𝑃𝑙𝑜𝑎𝑑 − ෍ 𝑃𝑖𝑡 𝑈𝑖𝑡 )
𝑑𝜆 𝑑𝜆 𝑑𝜆
𝑡=1 𝑖=1

𝑡
=σ𝑇𝑡=1 𝜆𝑡 (𝑃𝑙𝑜𝑎𝑑 − σ𝑁 𝑡 𝑡
𝑖=1 𝑖 𝑈𝑖 ) = [170,520,500,330]
𝑃

⇒ ⇒ [3.4, 10.4, 16, 6.6]

10:26 AM 22
3rd Iteration [Step 2]
Repeat the same steps

Economic Dispatch (ED)

Solve for 𝑃1 , 𝑃2 , 𝑃3 and λ , 𝑤𝑒𝑔𝑒𝑡
𝑑 𝑃1 =136.36
F P = 10 + 0.004P1 = λ1 𝑃2 =509.09 > limit
𝑑𝑃1 1 1
𝑑 𝑃3 =454.54 > limit
F P = 8 + 0.005P2 = λ2 λ=10.54
𝑑𝑃2 2 2
𝑑
F P = 6 + 0.01P3 = λ3 Set 𝑃2 , 𝑃3 at max limits, giving
𝑑𝑃3 3 3
𝑃1 + 𝑃2 + 𝑃3 =1100 𝑃1 =500
𝑃2 =400
Assume all units are operating within limits so λ1 =λ2 =λ3 =λ 𝑃3 =200
2:37 PM 23
10:26 AM 24
10:26 AM 25
Reading
• D. Bertsekas, G. Lauer, N. Sandell and T. Posbergh, "Optimal short-
term scheduling of large-scale power systems," in IEEE Transactions
on Automatic Control, vol. 28, no. 1, pp. 1-11, Jan 1983

10:12 AM 26
10:26 AM 27
10:26 AM 28
10:26 AM 29
Assignment 2
• Due on 17/11/2017 (11:50 am)

10:26 AM 30