ADVANCED POWER

SYSTEM OPTIMIZATION
(ELL776)

-DR YASHASVI BANSAL

ASSISTANT PROFESSOR
DEPARTMENT OF ELECTRICAL ENGINEERING
IIT DELHI-110016
 UNIT 1:
INTRODUCTION TO
OPTIMIZATION AND ITS
APPLICATION IN POWER
SYSTEMS
 3

CONTENTS
Introduction to optimization
History of optimization
Basic concepts in optimization
Overview of optimization in power systems

© Dr Yashasvi Bansal, IIT Delhi, 2025

 4
What is Optimization?
Optimization process aims to minimize or maximize a certain objective to achieve most
optimal but feasible solution.

s. t.

© Dr Yashasvi Bansal, IIT Delhi, 2025 Source: Prof. Abhyankar, IIT Delhi, 2025
 5

EXAMPLE

( )
min x − 5x + 4
2

2.5

-2.25

© Dr Yashasvi Bansal, IIT Delhi, 2025

 6

EXAMPLE

© Dr Yashasvi Bansal, IIT Delhi, 2025

 7

EXAMPLE

(
min x − 5x + 4
2
)
S. t.
2.5 4

-2.25
x − 5x + 4  0
2

x  2.5

© Dr Yashasvi Bansal, IIT Delhi, 2025

 8

EXAMPLE

© Dr Yashasvi Bansal, IIT Delhi, 2025

 9
ESSENTIAL INGREDIENTS OF OPTIMIZATION PROBLEM

Almost every practical optimization problem contains three essential contents:

• At least one objective function to be (minimized / maximized):

1. Objective function: The objective is to maximize or minimize a mathematical function.

• Different limits imposed on the optimization process are called as constraints.

2. Equality constraints

3. Inequality constraints

Region bounded by the constraints is known as feasible region for the optimization process.

© Dr Yashasvi Bansal, IIT Delhi, 2025

 FEW OPERATIONS ON OBJECTIVE FUNCTION 10

© Dr Yashasvi Bansal, IIT Delhi, 2025

 11

FEASIBLE REGION
Feasible Region

X2 Non-linear inequality
constraint
Linear inequality
Non-linear inequality
constraint
constraint
linear inequality
constraint

Non-linear inequality
constraint

X1
linear inequality constraint
© Dr Yashasvi Bansal, IIT Delhi, 2025
 12

BLOCK DIAGRAM OF OPTIMIZATION

SOLVER

Objective Function Outcome for variables

Optimization
Equality Constraints Solver
Lagrange Multipliers
Inequality Constraints

© Dr Yashasvi Bansal, IIT Delhi, 2025

 13

EXAMPLE

© Dr Yashasvi Bansal, IIT Delhi, 2025

 POPULAR PACKAGES 14

◼ Generalized Algebraic Modeling System (GAMS)

◼ The GAMS is a high-level modeling system for mathematical optimization. GAMS is
designed for modeling and solving linear, nonlinear, and mixed-integer optimization
problems.

◼ A Mathematical Programming Language (AMPL)

◼ AMPL is a high-level language for describing and solving mathematical optimization
problems. Unlike general programming languages such as Python, C++, or Java, which are
designed for a broad range of applications, AMPL is dedicated to optimization modeling.

◼ MATLAB Optimization Toolbox

◼ Optimization Toolbox is software that solves linear, quadratic, conic, integer, multi-
objective, and nonlinear optimization problems.

© Dr Yashasvi Bansal, IIT Delhi, 2025

 HISTORICAL DEVELOPMENTS 15

◼ Antiquity
❑ 300 BC – Euclid – minimal distance between points is line

◼ 17th and 18th Century: Birth of calculus

❑ 1660-1670 - Newton, Leibniz – Calculus – foundation for optimization

techniques
❑ 1740-1755 – Euler, Lagrange – Calculus of variations

◼ 19th Century: Pinoeering Methods

❑ 1824 – Fourier – attempt to solve linear problem

❑ 1847 – Cauchy – gradient method

◼ 20th Century: The Foundations of Modern Optimization

❑ 1902 – Farkas – Lemma, foundation for KKT conditions
❑ 1917 – Hancock – First text book on optimization – Theory of maxima and minima
❑ 1939 – Kantrovich – LP model and algorithm (Nobel prize in economics in 1975)
© Dr Yashasvi Bansal, IIT Delhi, 2025
 HISTORICAL DEVELOPMENTS 16

◼ Post World War II: The Boom of Optimization

❑ 1944 – J Von Neuman – Dynamic Programming

❑ 1947 – Dantzig – Simplex method for LP

❑ 1947 – J Von Neuman – Theory of Duality

◼ 1950 Onwards: Modern Optimization Techniques

❑ 1951 – Kunh Tucker – Optimality conditions for non-linear problem

❑ 1957 – Bellman – Optimality principle

❑ 1984 – N Karmarkar – polynomial time algorithm for LP methods

(Interior point method)
❑ 1985 – Meta-heuristic techniques (GA proposed by Smith in 1980)

© Dr Yashasvi Bansal, IIT Delhi, 2025

 17

CLASSIFICATION OF
OPTIMIZATION PROBLEMS

© Dr Yashasvi Bansal, IIT Delhi, 2025

 CLASSIFICATION OF OPTIMIZATION PROBLEMS
18

◼ Based on existence of constraints

◼ Based on nature of variables

◼ Based on nature of equations

◼ Based on permissible values of design variables

◼ Deterministic / stochastic

◼ Single objective / multi-objective

© Dr Yashasvi Bansal, IIT Delhi, 2025
 CLASSIFICATION BASED ON EXISTENCE OF CONSTRAINTS 19

Unconstrained Optimization Problem

Constrained Optimization Problem

where X is an n-dimensional vector called the design vector, f (X) is termed the objective function, and gj (X) and lj
(X) are known as inequality and equality constraints, respectively. The number of variables n and the number of
constraints m and/or p need not be related in any way.
© Dr Yashasvi Bansal, IIT Delhi, 2025
 20

(
min x2 − 5x + 4 ) Unconstrained optimization

(
min x − 5x + 4
2
) Constrained optimization
s.t.
x2 − 5x + 4  0
x  2.5

© Dr Yashasvi Bansal, IIT Delhi, 2025

 CLASSIFICATION BASED ON NATURE OF VARIABLES 21

Parameter or Static optimization problem

• Problem is to find values to a set of design parameters that make some prescribed
function of these parameters minimum subject to certain constraints.
• Variables have numerical values, fixed with respect to time or any other parameter

Trajectory or Dynamic Optimization Problem

• Objective is to find a set of design parameters, which are all continuous functions
of some other parameter, that minimizes an objective function subject to a set of
constraints.
• Variables are functions of time or some other parameter

© Dr Yashasvi Bansal, IIT Delhi, 2025

 22

Parameter or Static optimization problem

Trajectory or Dynamic Optimization Problem

Design variables are functions of the length

© Dr Yashasvi Bansal, IIT Delhi, 2025 parameter t
 23
CLASSIFICATION BASED ON NATURE OF EQUATIONS

◼ Linear

◼ Non-linear

◼ Geometric

◼ Quadratic

© Dr Yashasvi Bansal, IIT Delhi, 2025

 24

CLASSIFICATION BASED ON NATURE OF EQUATIONS: LINEAR

PROGRAMMING PROBLEM (LP)
Objective function and all constraints are linear functions of variable

min

s.t.

© Dr Yashasvi Bansal, IIT Delhi, 2025

 25

CLASSIFICATION BASED ON NATURE OF EQUATIONS: NON-

LINEAR PROGRAMMING PROBLEM (NLP)

Any of the functions among objective or constraints is non-linear functions of variable

© Dr Yashasvi Bansal, IIT Delhi, 2025

 26

CLASSIFICATION BASED ON NATURE OF EQUATIONS: GEOMETRIC

PROGRAMMING PROBLEM
Objective function and constraints are expressed as posynomials in X

A function h(X) is posynomial if h is expressed as sum of power terms of form:

A posynomial with N terms is written as:

© Dr Yashasvi Bansal, IIT Delhi, 2025

 27

min

s.t.

© Dr Yashasvi Bansal, IIT Delhi, 2025

 28

CLASSIFICATION BASED ON NATURE OF EQUATIONS: QUADRATIC

PROBLEM

Objective function is quadratic and all constraints are linear

min

s.t.

© Dr Yashasvi Bansal, IIT Delhi, 2025

 29

CLASSIFICATION BASED ON PERMISSIBLE VALUES OF

DESIGN VARIABLES

Real Valued Programming Problem

Variables can take any real number, subject to constraints

Integer Programming Problem

Variables can take only integer values, subject to constraints

Binary Integer Programming Problem

Variables can take only values of 1 or 0

© Dr Yashasvi Bansal, IIT Delhi, 2025

 30
EXAMPLE

© Dr Yashasvi Bansal, IIT Delhi, 2025

 31

CLASSIFICATION BASED ON THE DETERMINISTIC NATURE OF THE

VARIABLES

Deterministic Stochastic
◼ All variables are
◼ Some or all variables
deterministic in nature and predefined
parameters are
probabilistic in nature

© Dr Yashasvi Bansal, IIT Delhi, 2025

 32

CLASSIFICATION BASED ON THE NUMBER OF OBJECTIVE FUNCTIONS

Single Objective Multi-objective

◼ One objective to be ◼ Combination of a number of objectives
minimized ◼ Objectives can be combined
◼ Example: minimize cost through various means
◼ Example: minimize cost and maximize
reliability

© Dr Yashasvi Bansal, IIT Delhi, 2025

 33
EXAMPLE
Pareto Optimal Solutions: Solutions that
cannot be improved in any of the
objectives without degrading at least one
of the other objectives.

The edge between the lowest values of f1 and f2 is called the Pareto Front. Any point on this front is
considered “Pareto optimal”. By moving along the curve, you could minimize cost at the expense of time, or
minimize time at the expense of cost, but you can not improve both at once.

© Dr Yashasvi Bansal, IIT Delhi, 2025

 34

WHAT CLASS OF PROBLEM IS USED IN POWER

SYSTEMS, MOST OF THE TIME?

Power System Optimization

NLP Constrained Static Real + Integer Deterministic Single

Objective

The other forms may be used for certain specific purposes!

© Dr Yashasvi Bansal, IIT Delhi, 2025
 35

A FEW IMPORTANT CONCEPTS IN OPTIMIZATION

© Dr Yashasvi Bansal, IIT Delhi, 2025

 36
CONTINUOUS AND CONTINUOUSLY
DIFFERENTIABLE FUNCTION
IF F(X) IS CONTINUOUS AT EVERY POINT IN A REGION R, THEN F(X) IS SAID TO BE CONTINUOUS
THROUGHOUT R.

f(x) f(x) is discontinuous.

f(x) is continuous, but

f(x)
f (x)  df (x) is not.
dx
x
A function f is said to be continuously differentiable if the derivative f′(x) exists and is itself a continuous function.
© Dr Yashasvi Bansal, IIT Delhi, 2025
 37

CONVEXITY

If we can prove that a minimization problem is convex:

• Convex feasible set

• Convex objective function

→ Then, the problem has one and only one solution i.e., there
is only one minimum and it is thus the global minimum

© Dr Yashasvi Bansal, IIT Delhi, 2025

 38

LOCAL AND GLOBAL MINIMUM

© Dr Yashasvi Bansal, IIT Delhi, 2025

 Convex Function
A function f (X) is said to be convex if for any pair of points

that is, if the segment joining the two points lies entirely above or on the graph of f (X).

A function f (X) is strictly convex or concave if the strict inequality holds for any X1 = X2.

© Dr Yashasvi Bansal, IIT Delhi, 2025

 40
Functions of one variable

(a) convex function in one variable; (b) concave function in one variable.

© Dr Yashasvi Bansal, IIT Delhi, 2025

 41
Functions of two variable

(a) convex function in one variables; (b) concave function in two variables.

© Dr Yashasvi Bansal, IIT Delhi, 2025

 42

Function that is convex over certain region and concave over certain other region.

© Dr Yashasvi Bansal, IIT Delhi, 2025

 43

2 f
I f f  (x) =  0 t h e n f (x) is c oncave.
x 2

2 f
I f f  (x) = 2  0 t h e n f (x) is c onvex.
x
f(x) f(x)

f (x)  0

f (x)  0

x x
xa xb xa xb
Concave Function Convex Function

© Dr Yashasvi Bansal, IIT Delhi, 2025

 PROBLEM 44

Indicate whether it is a convex function or non-convex function?

© Dr Yashasvi Bansal, IIT Delhi, 2025

 45

CONVEX SET

© Dr Yashasvi Bansal, IIT Delhi, 2025

 46

EXAMPLE OF CONVEX SET

© Dr Yashasvi Bansal, IIT Delhi, 2025

 47
EXAMPLE OF NON-CONVEX SET
 48

PROBLEM
What does the set {(x, y) | x2 + y2 ≤ 4} represent geometrically. Graph
the set and indicate whether it is a convex set or not?

© Dr Yashasvi Bansal, IIT Delhi, 2025

 49

PROBLEM
Sketch a graph for each of the following and indicate whether the set is convex?

© Dr Yashasvi Bansal, IIT Delhi, 2025

 TEST FOR CONVEXITY FOR MULTI- 50

DIMENSIONAL SYSTEM
Look for Hessian Matrix!

Set is strictly convex if H(x) is positive definite, meaning all eigenvalues of H(x)
are greater than zero

Set is convex if H(x) is positive semi-definite, meaning eigenvalues of H(x) are

greater than or equal to zero

© Dr Yashasvi Bansal, IIT Delhi, 2025

 51
EXAMPLE

© Dr Yashasvi Bansal, IIT Delhi, 2025

 52
NECESSARY AND SUFFICIENT CONDITIONS

A condition N is necessary for a result R if R can be true only if N is

true.
RN

A condition S is sufficient for a result R if R is true if S is true

(if R is true guarantees that S is also true.)
SR

A condition T is necessary and sufficient for a result R iff T is true.

TR
© Dr Yashasvi Bansal, IIT Delhi, 2025
 53

EXAMPLE
• Statement 1: “you sit the exam” and Statement 2: “you pass the exam”.
Statement 1 is ….............for statement 2.
• Statement 1: “you achieve an overall grade of over 70% in all of the modules
that you have studied as part of your economics degree” and Statement 2: “you
get a first-class degree in economics”
Statement 1 is ….............for statement 2.
• Statement 1: “I am a male sibling” and Statement 2: “I am a brother”.
Statement 1 is ….............for statement 2.

© Dr Yashasvi Bansal, IIT Delhi, 2025

 54

NECESSARY AND SUFFICIENT CONDITIONS

df = 0
Necessary Condition:
dx f(x)

Sufficient Condition: f (x)  0

d2 f
For Minima: 0
2
dx
A
d2 f
For Maxima: 0
2
xb x
dx xa df = 0
dx

© Dr Yashasvi Bansal, IIT Delhi, 2025

 55

EXAMPLE

© Dr Yashasvi Bansal, IIT Delhi, 2025

 56

EXAMPLE

© Dr Yashasvi Bansal, IIT Delhi, 2025

 57
NECESSARY CONDITIONS FOR MULTI-
DIMENSIONAL CASES
At maximum or minimum value of f (x1, x2 , xn )

 f 
 x 
 1
 f 
x  = 0
 2
 
 f 
 
xn 
A point where this condition is satisfied is a stationary point – may be maxima,
may be minima!

© Dr Yashasvi Bansal, IIT Delhi, 2025

 58
SUFFICIENT CONDITIONS FOR MULTI-
DIMENSIONAL CASES
Calculate the Hessian Matrix at stationary point

 2 f 2 f 2 f 
 x2 x1x2 x1xn 
 1

 2 f 2 f 2 f 
 
 x 2 x1 x22 x2 xn 

 2 f 

 x1x2 
 2 
  f 2 f 2 f 
 x n x1 xn x2 x2n 

© Dr Yashasvi Bansal, IIT Delhi, 2025

 59
SUFFICIENT CONDITIONS FOR MULTI-
DIMENSIONAL CASES

Calculate the eigenvalues of the Hessian matrix at the stationary point

❑ If all the eigenvalues are greater or equal to zero:
▪ The matrix is positive semi-definite
▪ The stationary point is a minimum
❑ If all the eigenvalues are less or equal to zero:
▪ The matrix is negative semi-definite
▪ The stationary point is a maximum
❑ If some or the eigenvalues are positive and other are negative:
▪ The stationary point is a saddle point

© Dr Yashasvi Bansal, IIT Delhi, 2025

 60

© Dr Yashasvi Bansal, IIT Delhi, 2025

 61

INDEFINITENESS OF A MATRIX

H is indefinite and

© Dr Yashasvi Bansal, IIT Delhi, 2025

 62
EXAMPLE

© Dr Yashasvi Bansal, IIT Delhi, 2025

 63

© Dr Yashasvi Bansal, IIT Delhi, 2025

 64
PROBLEM

© Dr Yashasvi Bansal, IIT Delhi, 2025

 65
PROBLEM

For f(x, y) = 4x + 2y - x2 -3y2

a) Find the gradient. Use that to find a stationary point(x, y) that makes
the gradient 0.
b) Use the eigenvalues of the Hessian at that point to determine whether
the stationary point in a) is a maximum, minimum, or neither.

© Dr Yashasvi Bansal, IIT Delhi, 2025

 66

PROBLEM

© Dr Yashasvi Bansal, IIT Delhi, 2025

 67
INTERIOR AND BOUNDARY CONDITIONS

© Dr Yashasvi Bansal, IIT Delhi, 2025

 68

CONTOURS
 69

CONTOURS

© Dr Yashasvi Bansal, IIT Delhi, 2025

 70

CLASSICAL TECHNIQUES: BROAD CLASSIFICATION

Analytical Methods Numerical Methods

By and large involve differential Direct Search: No derivatives

calculus. Effectively boils down (Fibonacci series)
to solving a set of linear / non-
linear algebraic equations Descent Methods: First /
Second derivatives (Steepest
descent or Cauchy, Newton’s
method)

© Dr Yashasvi Bansal, IIT Delhi, 2025

 71

CLASSICAL TECHNIQUES: PROBLEM-WISE CLASSIFICATION

Linear Programming Non-linear Programming Integer Programming

Simplex Method Newton’s Method Branch and

and its variants and variants Bound Method

Interior Point Steepest Descent

Method (Cauchy) Method

Augmented
Lagrange
Multiplier Method

© Dr Yashasvi Bansal, IIT Delhi, 2025

 72
CLASSICAL TECHNIQUES: NLP

Direct Methods (Constraints Indirect Methods (Sequence

handled in explicit manner) of unconstrained min.
problems)

SLP (First order Augmented

Taylor series Lagrange
expansion) Multiplier Method

SQP Penalty Function

Method

Generalized
Reduced Gradient

© Dr Yashasvi Bansal, IIT Delhi, 2025

 73

OPTIMIZATION IN POWER
SYSTEMS

© Dr Yashasvi Bansal, IIT Delhi, 2025

 74

TIMELINE OF ACTIVITIES
◼ Long term planning
5-10 years or more
❑ Transmission expansion planning
❑ Generation expansion planning
◼ Short term planning
❑ Scheduling (Unit Commitment) 3 months to 1 week
❑ Optimal Distribution system network topology
❑ Short term market
◼ Operational planning
❑ Scheduling (Unit Commitment, ELD) 1 week to 1 day
❑ Day-ahead market ahead
❑ Security
◼ Operations
❑ Dispatch Hour ahead to real
❑ Security time
❑ Real time market
❑ Monitoring (State Estimation)

© Dr Yashasvi Bansal, IIT Delhi, 2025

 FLOWCHART FOR ANY OPTIMIZATION 75
PROBLEM

© Dr Yashasvi Bansal, IIT Delhi, 2025

 76

MATLAB: OPTIMIZATION TOOLBOX

© Dr Yashasvi Bansal, IIT Delhi, 2025

 MATLAB PROGRAMS OR FUNCTIONS FOR SOLVING 77
OPTIMIZATION PROBLEMS

© Dr Yashasvi Bansal, IIT Delhi, 2025

 78

© Dr Yashasvi Bansal, IIT Delhi, 2025

 79

STEPS
• Step 1: Formulating the optimization problem in the format required by the
MATLAB program
▪ the objective function for minimization
▪ the constraints in “≤” form with zero or constant value on the righthand
side of the inequalities.
• Step 2: involves writing an m-file for the objective function.
• Step 3: involves writing an m-file for the constraints.
• Step 4: involves setting the various parameters at proper values depending on
the characteristics of the problem and the desired output.

© Dr Yashasvi Bansal, IIT Delhi, 2025

 80

BASIC MATLAB EXAMPLE

Given fun1(x) = x3 - 2x + cos(x)
1. Create .m file
2. Write in command window: fun(3*pi/2)

Given fun2(x)= x2 +2x+3 = 0

A function can also be evaluated at some given arguments (input parameters) via the feval command.
[x1, x2] = feval('fun2',1,2,3)
 81
MATLAB PROBLEM
Find the solution of given optimization problem using the MATLAB function
"linprog".

© Dr Yashasvi Bansal, IIT Delhi, 2025

 82
MATLAB PROBLEM
Find the solution of given optimization problem using the MATLAB function
"linprog".

© Dr Yashasvi Bansal, IIT Delhi, 2025

 83
MATLAB PROBLEM
Find the solution of given optimization problem using the MATLAB function
"fmincon".

© Dr Yashasvi Bansal, IIT Delhi, 2025

 THANK YOU

For any queries, please email at [email protected]

© Dr Yashasvi Bansal, IIT Delhi, 2025