JMEE7311

Power Systems Operation

and Management
Second lecture
 Mathematical methods and tools applied to power system
operation ( Optimization methods,…).

Dr. Nassim Iqteit

23-2-2019
Daniel Kirschen
1
Introduction to Optimization
Where can be used ?
 Minimize cost of building a transformer

 Minimize cost of supplying power

 Minimize losses in a power system Most engineering activities have an objective:

 Maximize profit from a bidding strategy  Achieve the best possible design

 Economic dispatch problem  Achieve the most economical operating conditions

 Optimal Power Flow

 Unit Commitment

 Minimizing error (estate estimation)

2
 Decision Variables
 Objective function

F = f ( x1 , x2 , x3 , .. xn ) • Examples of decision variables:

– Dimensions of the transformer
decision variables – Output of generating units, position of taps
– Parameters of bids for selling electrical energy

 Optimization Problem • What value should the decision variables

take so that F = f ( x1, x2 , x3 , .. xn ) is minimum
or maximum?

 Minimization and Maximization

f(x*) f(x)
• If x = x* maximizes f(x) then it minimizes - f(x)

• maximizing f(x) is thus the same thing as x* x

minimizing g(x) = -f(x)

-f(x*) -f(x)
3
  Necessary Condition for Optimality (one dimension)
If x = x * maximises f ( x ) then:
df
df f(x) =0
f ( x ) < f ( x ) for x < x Þ
* *
> 0 for x < x * dx
dx f(x*)
df df df
f ( x ) < f ( x ) for x > x Þ
*
< 0 for x > x *
*

dx >0 <0
dx dx
df
If x = x maximises f ( x ) then
*
= 0 for x = x *
dx x* x
Example
• A, B, C, D are stationary points • B is a minimum
For what values of x is ? • A and D are maxima • C is an inflexion point
In other words, for what values of x is the
necessary condition for optimality satisfied? f(x)
d2 f
For x = A and x = D, we have: < 0 The objective function is concave around a maximum
dx 2
d2 f
For x = B we have: > 0 The objective function is convex around a minimum
dx 2
d2 f A B C D x
For x = C , we have: = 0 The objective function is flat around an inflexion point 4
dx 2
Necessary and Sufficient Conditions of Optimality

df
• Necessary condition: =0
dx
• Sufficient condition:
– For a maximum: d 2
f
<0
dx 2

d2 f
– For a minimum: >0
dx 2

5
  Feasible Set

 The values that the decision variables can take are usually limited
Examples:
• Physical dimensions of a transformer must be positive
• Active power output of a generator may be limited to a certain range (e.g. 200 MW to 500 MW)
• Reactive power output of a generator may be limited to a certain range (e.g. -100 MVAr to 150 MVAr)

 Interior and Boundary Solutions

 The values of the objective function outside
the feasible set do not matter f(x)

f(x) • A and D are interior maxima

• B and E are interior minima
• XMIN is a boundary minimum
• XMAX is a boundary maximum
Do not satisfy the
Optimality conditions!
xMINA D xMAX x xMIN A xMAX x
B D E
6
Feasible Set
 Two-Dimensional Case  Sufficient Conditions for Optimality
 Necessary Conditions for Optimality f(x1,x2)
f(x1,x2) maximum

¶f ( x 1 ,x 2 )
=0
¶x 1
x * ,x *
1 2
¶f ( x 1 ,x 2 )
=0
x1* ¶x 2 x * ,x * minimum
x1 1 2
x1
x2 *

x2 x2
f(x1,x2)
 Multi-Dimensional Case Saddle point
At a maximum or minimum value of f ( x1, x2 , x3 , .. xn )
we must have: ¶f = 0
¶x 1
¶f
=0 A point where these conditions are x1
¶x 2
satisfied is called a stationary point
x2
¶f
=0 7
¶x n
 Contours

• A contour is the locus of all the point that give the same value to the objective function

f(x1,x2)
x2
F2

F1

x1 x1
F1 F Minimum or maximum
2
x2

8
 Sufficient Conditions for Optimality

Calculate the Hessian matrix at the stationary point: • Calculate the eigenvalues of the Hessian matrix at the
 2 f 2 f 2 f  stationary point
  • If all the eigenvalues are greater or equal to zero:
 x12 x1x2 x1xn  – The matrix is positive semi-definite
  – The stationary point is a minimum
 2 f 2 f 2 f 
• If all the eigenvalues are less or equal to zero:
H  x2 x1 x22 x2 xn 
  – The matrix is negative semi-definite
  – The stationary point is a maximum
  • If some or the eigenvalues are positive and other are
 2 f 2 f  f 
2
negative:
 xn x1 xn x2 xn2 
 – The stationary point is a saddle point

 Eigenvalues first. If . 𝑯 − 𝜆𝑰 𝑥 = 0 has a nonzero solution, 𝑯 − 𝜆𝑰 is not invertible. The

determinant of 𝑯 − 𝜆𝑰 must be zero. This is how to recognize an eigenvalue :
Eigenvalues (): 𝐝𝐞𝐭(𝑯 − 𝜆𝑰) = 0 Syntax
Matlab • H=hessian(f,v)
 For each solve . 𝑯 − 𝜆𝑰 𝑥 = 0 or 𝑯𝑥 = 𝜆𝑥 to find an eigenvector x • e = eig(H) 9
  Calculation examples
must be positive definite (i.e. all eigenvalues must be positive)
1

Minimise C  x12  4 x22  2 x1 x2  2 2

Necessary conditions for optimality:  0   2  10  12  0
2  8
C
 2 x1  2 x2  0
x1   =5  13  0 The stationary point
ìx = 0 1 is a minimum
C
 2 x1  8 x2  0
í
x2 îx 2 = 0 syms x1 x2
C=x1^2 + 4*x2^2 - 2*x1*x2;
is a stationary point x2 H=hessian(C,[x1,x2]);
e = eig(H);
Sufficient conditions for optimality:
H=
C=9
  2C  2C  C=4
[ 2, -2]
  C=1 [ -2, 8]
 x1 x1 x2   2 2  Matlab
2

   2 8 
H=  e=

  2
C  C  
2
 x1 5 - 13^(1/2)
13^(1/2) + 5
 x x x22 
 2 1 The stationary point
Minimum: C=0 is a minimum
10
 2 2 2
 0   2  4  8  0
2  6
Minimize C = -x12 + 3x22 + 2x1 x2   =2  2 5  0
The stationary point
or  =2  2 5  0 is a saddle point
Necessary conditions for optimality:
¶C
= -2x1 + 2x2 = 0
¶x1 ì x1 =0 x2

¶C í C=0

¶x2
= 2x1 + 6x2 = 0 îx 2 = 0 C=9 syms x1 x2
C=-x1^2 + 3*x2^2 + 2*x1*x2;
is a stationary point C=4 H=hessian(C,[x1,x2]);
e = eig(H);
Sufficient conditions for optimality: C=1

C=-4 C=-9
H=
C=-9 C=-4 C=-1

  2C  2C  x1 [ -2, 2]
  C=1 Matlab
[ 2, 6]
 x1 x1 x2   2 2 
2
H=     2 6
C=4 e=

  2
C  C  
2
 C=9 2 - 2*5^(1/2)
2*5^(1/2) + 2
 x x x22  C=0
The stationary point
 2 1 This stationary point is a saddle point is a saddle point

11
Optimization with Constraints
 Optimization with Equality Constraints
• There are usually restrictions on the values that the decision variables can take

Minimise
Number of Constraints
f ( x1 , x2 ,.. , xn ) Objective function • N decision variables
• M equality constraints
subject to: • If M > N, the problems is over-constrained
w 1 ( x1 , x2 ,.. , xn ) = 0 – There is usually no solution
• If M = N, the problem is determined
Equality constraints – There may be a solution

w m ( x1 , x2 ,.. , xn ) = 0
• If M < N, the problem is under-constrained
– There is usually room for optimization

12
 Example A Example B
Solution by substitution
Minimise f ( x 1 , x 2 ) = 0.25 x 12 + x 22 Minimise C = a1 + a 2 + b1 x 12 + b 2 x 22
Subject to w ( x 1 , x 2 ) º 5 - x 1 - x 2 = 0
Subject to: x 1 + x 2 = L
Þ x 2 = L - x1
x2
Þ C = a1 + a 2 + b1 x 12 + b 2 ( L - x 1 )
2

w ( x 1 ,x 2 ) º 5 - x 1 - x 2 = 0 • Difficult
dC
 2b1 x1  2b2  L  x1   0 • Usually impossible
dx1 when constraints are
non-linear
Minimum b2 L  b1 L 
 x1    x2   • Provides little or no
b1  b2  b1  b2  insight into solution
d 2C
x1
 2b1  2b2  0  minimum • Solution using
f ( x 1 ,x 2 ) = 0.25 x + x
2
1
2
2
dx12 Lagrange multipliers
13
 Gradient
Consider a function f (x1 , x2 ,.. , xn ) Properties of the Gradient
æ ¶f ö • Each component of the gradient vector indicates the
ç ¶x1 ÷ rate of change of the function in that direction
ç ÷
ç ¶f ÷ • The gradient indicates the direction in which a
function of several variables increases most rapidly
The gradient of f is the vector Ñf = ç ¶x2 ÷÷
ç • The magnitude and direction of the gradient usually
ç ÷ depend on the point considered
ç ¶f ÷
ç ÷ • At each point, the gradient is perpendicular to the
çè ¶xn ÷ø contour of the function
f = f3
Example A Example B

f ( x , y ) = ax 2 + by 2 f ( x , y ) = ax + by f = f2 y Ñf

y
æ ¶f ö æ ¶f ö f = f1 Ñf

ç ¶x ÷ æ 2 ax ö ç ¶x ÷ æ a ö
Ñf = ç ÷ = ç
B
Ñf = ç ÷ = ç ÷ Ñf
÷ ¶f
¶f è 2 by ø C ç ÷ è bø
ç ÷ A
è ¶y ø
è ¶y ø x
x
D 14
 Lagrange multipliers æ ¶w
ç ¶x 1
ö
÷
Minimise f  x1 , x2   0.25 x12  x22 , subject to   x1 , x2   5  x1  x2  0 Ñw = ç ÷
çç ¶w ÷÷
w ( x1 , x 2 ) = 5 - x1 - x 2 w ( x1 , x 2 ) è ¶x 2 ø
Ñf
f ( x 1 ,x 2 ) = 6 f ( x1 , x 2 ) = 6
f = 0.25 x 12 + x 22 = 6 f ( x1 , x 2 ) = 5 Ñf f ( x1 , x 2 ) = 5
æ ¶f
Ñw
f = 0.25 x 12 + x 22 = 5 ö Ñf Ñw
ç ¶x 1 ÷
2 Ñf = ç ÷ 3
1 ç ¶f ÷
ç
è ¶x 2
÷
ø Ñw

w ( x1 , x 2 ) • We stop when the gradient of the function

4 Ñf w ( x 1 ,x 2 ) is perpendicular to the constraint because

f ( x 1 ,x 2 ) = 6 5 Ñf moving further would increase the value

of the function
A f ( x 1 ,x 2 ) = 6
f ( x 1 ,x 2 ) = 5 f ( x 1 ,x 2 ) = 5
A Ñf
? Ñf
Ñw C Ñf
The solution must be on the constraint B Ñw
To reduce the value of f, we must move At the optimum, the gradient of the B
in a direction opposite to the gradient function is parallel to the gradient Ñw 15
of the constraint Ñf + l Ñw = 0
At the optimum, we must have: Ñf Ñw
Which can be expressed as: Ñf + l Ñw = 0

In terms of the co-ordinates: ¶f ¶w

+l =0
¶x 1 ¶x 1
¶f ¶w
+l =0
¶x 2 ¶x 2

The constraint must also be satisfied: w ( x1 , x 2 ) = 0

l is called the Lagrange multiplier

16
Lagrangian function
To simplify the writing of the conditions for optimality,
it is useful to define the Lagrangian function:

( x 1 , x 2 ,l ) = f ( x 1 , x 2 ) + lw ( x 1 , x 2 )
The necessary conditions for optimality are then given
by the partial derivatives of the Lagrangian:

  x1 , x2 ,   f 
  0
x1 x1 x1
  x1 , x2 ,   f 
  0
x2 x2 x2
  x1 , x2 ,  
   x1 , x2   0
 17
Generalization
Minimise
f  x1 , x2 ,.. , xn 
subject to:
1  x1 , x2 ,.. , xn   0

m  x1 , x2 ,.. , xn   0
Lagrangian:

= f  x1 ,.. , xn   11  x1 ,.. , xn    mm  x1 ,.. , xn 

• One Lagrange multiplier for each constraint
• n + m variables: x1, …, xn and λ1, …, λm
18
Optimality conditions
= f  x1 ,.. , xn   11  x1 ,.. , xn    mm  x1 ,.. , xn 
 f 1 m
  1   m 0
x1 x1 x1 x1
n equations
 f 1 m
  1   m 0
xn xn xn xn

 1  x1 , , xn   0
1
m equations

 m  x1 , , xn   0 n + m equations in
m n + m variables
19
Example
Minimise f ( x 1 , x 2 ) = 0.25 x 12 + x 22 subject to w ( x 1 , x 2 ) º 5 - x 1 - x 2 = 0

 x1 , x2 ,    0.25 x12  x22    5  x1  x2 

x2
  x1 , x2 ,  
 0.5 x1    0 w ( x1 , x 2 ) º 5 - x1 - x 2 = 0
x1  x1  2
  x1 , x2 ,  
 2 x2    0 1
x2  x2   Minimum
2 1
  x1 , x2 ,   f ( x1 , x 2 ) = 5
 5  x1  x2  0 1
  5  2    0 4 x1
2
 2
 x1  4
 x2  1 20
 Optimization with inequality constraints

Minimise
f ( x1 , x2 ,.. , xn ) Objective function
subject to:
w 1 ( x1 , x2 ,.. , xn ) = 0
Equality constraints
w m ( x1 , x2 ,.. , xn ) = 0
and:
g1 ( x1 , x2 ,.. , xn ) £ 0
Inequality constraints
gp ( x1 , x2 ,.. , xn ) £ 0
21
 https://www.mathworks.com/help/optim/ug/fmincon.html
Optimization Toolbox
fmincon
• Nonlinear programming solver.

• b and beq are vectors, A and Aeq are matrices, c(x) and ceq(x) are functions that return vectors, and f(x) is
a function that returns a scalar. f(x), c(x), and ceq(x) can be nonlinear functions.

• x, lb, and ub can be passed as vectors or matrices.

Syntax: x = fmincon(fun,x0,A,b,Aeq,beq,lb,ub) [x,fval] = fmincon(fun,x0,A,b,Aeq,beq,lb,ub)

defines a set of lower and upper bounds on the design variables in x, so that the solution is always in the
range lb ≤ x ≤ ub. If no equalities exist, set Aeq = [] and beq = []. If x(i) is unbounded below, set lb(i) = -Inf, and
if x(i) is unbounded above, set ub(i) = Inf.

22
 Example A:
Example B:

𝒎𝒊𝒏 𝐶( 𝑥1 , 𝑥2 ) 𝒎𝒊𝒏 100(𝑥2 − 𝑥12 )2 +(1 − 𝑥1 )2

𝑥1 ,𝑥2 𝑥1 ,𝑥2
s.t
𝐶 = 𝑥12 + 4𝑥22 − 2𝑥1 , 𝑥2
𝑥1 +2𝑥2 ≤ 1
2𝑥1 +𝑥2 = 1
0 ≤ 𝑥1 ≤1
0 ≤ 𝑥2 ≤ 0.5
fun = @(x) x(1)^2 + 4*x(2)^2 - 2*x(1)*x(2) fun = @(x)100*(x(2)-x(1)^2)^2 + (1-x(1))^2;
A = [];
b = []; x=
A = [1,2];
Aeq = []; 0 0 x=
beq = []; b = [1];
0.4149 0.1701
lb = [-inf,-inf]; Aeq = [2,1];
fval =
ub = [inf,inf]; beq = [1];
0 fval =
x0 = [0,0]; lb = [0,0];
0.3427
[x,fval] = fmincon(fun,x0,A,b,Aeq,beq,lb,ub) ub = [1,0.5];
x0 = [0.5,0.25];
[x,fval] =
fmincon(fun,x0,A,b,Aeq,beq,lb,ub)

23
 Applications

Economic dispatch problem

24
 Economic dispatch problem

L
A B C

• Several generating units serving the load

• What share of the load should each generating unit produce?
• Consider the limits of the generating units
• Ignore the limits of the network

25
 Characteristics of the generating units
J/h
B T G

Input
Fuel Electric Power MW
Pmin Pmax
(Input) (Output)
Output
• Thermal generating units
• Consider the running costs only
• Input / Output curve
– Fuel vs. electric power
• Fuel consumption measured by its energy content
• Upper and lower limit on output of the generating unit

26
  Cost Curve  Incremental Cost Curve
• Multiply fuel input by fuel cost • Incremental cost curve
• No-load cost D FuelCost vs Power
– Cost of keeping the unit running if it could D Power
produce zero MW
• Derivative of the cost curve
• In $/MWh
$/h • Cost of the next MWh
Cost

Cost [$/h] Incremental Cost

[$/MWh]
No-load cost
MW
Pmin Pmax ∆F
Output ∆P
MW MW

27
 Mathematical formulation

• Objective function
C = CA (PA ) + CB (PB ) + CC (PC ) A B C
L

• Constraints
– Load / Generation balance:
L = PA + PB + PC
– Unit Constraints:
PAmin £ PA £ PAmax
PBmin £ PB £ PBmax
PCmin £ PC £ PCmax
This is an optimization problem ?
28
 Application to Economic Dispatch

x1 x2 minimise f  x1 , x2   C1  x1   C2  x2 
L

s.t.   x1 , x2   L  x1  x2  0
G1 G2

 x1 , x2 ,    C1  x1   C2  x2     L  x1  x2 
 dC1
   0
x1 dx1 dC1 dC2
 
 dC2 dx1 dx2
   0
x2 dx2
 Equal incremental cost
 L  x1  x2  0 solution

29
 Equal incremental cost solution

C1 ( x 1 ) C2 ( x 2 )
Cost curves:

x1 x2

dC 1 dC 2
Incremental dx 1 dx 2
cost curves:

x1 x2
30
 Interpretation of this solution

dC 1 dC 2
dx 1 dx 2

x1* x1 x2* x2

-
L -
+

If < 0, reduce λ
L-x -x *
1
*
2 If > 0, increase λ

31
 dC1
: Cost of one more MW from unit 1
 Physical interpretation dx1
dC2
: Cost of one more MW from unit 2
dC C
C( x )  lim dx2
dx x 0 x dC1 dC2
Suppose that 
DC dx1 dx2
Dx For x sufficiently small: dC1
Decrease output of unit 1 by 1MW  decrease in cost=
dC dx1
C   x
dx dC2
x Increase output of unit 2 by 1MW  increase in cost=
If x  1MW : dx2
dC(x) dC dC2 dC1
C  Net change in cost=  0
dx2 dx1
dx dx
It pays to increase the output of unit 2 and decrease the
The incremental cost is the cost of output of unit 1 until we have:
one additional MW for one hour. dC 1 dC 2
= =l
This cost depends on the output of dx 1 dx 2
the generator. The Lagrange multiplier λ is thus the cost of one more MW
at the optimal solution.
This is a very important result with many applications in
economics. 32
HW#1
Three generating units that have the following cost functions:

1
How should these units be dispatched if the generating units must supply a load of 350 MW
at minimum cost?
Solve this problem by Lagrangian function and by Optimization Toolbox fmincon.

2 Suppose a consumer has utility function 𝐶 𝑥1 , 𝑥2 = 𝐴𝑥1𝛼 𝑥21−𝛼 and faces the budget constraint
𝑝1 𝑥1 + 𝑝2 𝑥2 = 𝑀. 1) Find the values of demand functions 𝑥1 and 𝑥2 at minimum utility function 𝐶.

33
3 Solve this problem by Optimization Toolbox fmincon.

4 For the following function find:

𝑓 𝑥1 , 𝑥2 = 2𝑥13 + 𝑥1 𝑥22 + 5𝑥12 + 𝑥22
1) Find all stationary points of f(x1,x2).
2) Classify all stationary points (minimum, maximum or saddle points)

34
Thank you

35