Power System Planning

Lecture No. 2
Engr. Raheel Muzzammel
Department of Electrical Engineering
University of Lahore, Lahore, Pakistan

Engr. Raheel Muzzammel, University of Lahore

Review of Optimization Techniques in Power
System Planning

Engr. Raheel Muzzammel, University of Lahore

Problem Description
• Most of the operational and planning problems consist of the
following three major steps:
✓Definition
✓Modeling
✓Solution Algorithm
• Problem Definition
• In any optimization problem, the decision maker should decide on the
following items:
✓Decision (independent) and dependent variables
✓Constraints functions
✓Objective functions

Engr. Raheel Muzzammel, University of Lahore

Problem Description
• Decision and Dependent Variables
• Decision variables are the independent variables;
• The decision maker has to determine their optimum values and based on
those, other variables (dependent) can be determined.
• For instance,
• In an optimum generation scheduling problem, the active power generations of
power plants may be the decision variables.
• The dependent variables can be the total fuel consumption, system losses, etc. which
can be calculated upon determining the decision variables.
• In a capacitor allocation problem, the locations and the sizing of the capacitor banks
are the decision variables.
• The dependent variables may be bus voltages, system losses, etc.

Engr. Raheel Muzzammel, University of Lahore

Problem Description
• An n-decision variable
problem results in an n-
dimensional solution space in
which any point within that
space can be a solution.

Solution Space for

Two Dimensional
Space

Engr. Raheel Muzzammel, University of Lahore

Problem Description
• Constraints Functions
• In a real-life optimization problem, some limitations may apply to the solution space.
• These are typically technical, economical, environmental and similar limitations; named
as constraints which either directly or indirectly divide the solution space into acceptable
(feasible) and unacceptable (non-feasible) regions.
• The decision maker should find a solution point within the feasible region.
• For instance,
• In an optimum generation scheduling problem, the active power generations of the
power plants should be within their respective maximum and minimum values; or, the
total generation of the plants should satisfy total load and a specified reserve.
• In a capacitor allocation problem, a technical constraint may be the maximum number
of the capacitor banks which may be employed for a specific bus. An economical
constraint may be a limit on the total practical investment cost which should not be
violated.

Engr. Raheel Muzzammel, University of Lahore

Problem Description
• The way the constraints behave in
a two dimensional case is shown in
Figure.

Feasible and Non-Feasible

Region Due to Constraints

Engr. Raheel Muzzammel, University of Lahore

Problem Description
• Objective Function
• From the numerous points within the feasible region of a problem, the
decision maker should select the most desirable.
• The desirable should, however, be somehow defined.
• For instance,
• In a classroom, a teacher may select a student as the best if morality is the
main concern.
• He or she may select another if enthusiasm is observed.
• In fact, an objective function is a function in terms of the decision variables
by which the decision maker shows his or her desirable solution.

Engr. Raheel Muzzammel, University of Lahore

Problem Description
• In the Figure, If the objective function is defined
as maximizing x1, the solution ends up in point
A,
• If minimizing x2, point B would be the final
solution.
• In an optimum generation scheduling problem,
the objective function may be chosen as the
total fuel cost to be minimized.
• In a capacitor allocation problem, the objective
function may be the investment cost or the
system losses or both (to be minimized).
• The problem is considered to be single-objective
if just one objective function is to be optimized.
• It is in contrast to multi-objective optimization
problems in which several functions are to be
simultaneously, optimized.

Engr. Raheel Muzzammel, University of Lahore

Problem Description
• In a practical case, an optimization
problem may have many maximum and
minimum points.
• For instance,
• Consider the case depicted in Figure in
which the objective function is considered
to be a function of only x1 and is to be
maximized.
• As shown, there are some local optima in
the sense that they are optimum in the
• vicinity of nearby points.
• From those local optimum points, one is
the global optimum.

Engr. Raheel Muzzammel, University of Lahore

Problem Description
• Problem Modeling
• Once the decision variables, the constraints and the objective function terms are
decided, the decision maker should model the problem in a proper form to be
solved.
• The modeling depends much on the available tools and the algorithms for the
problem solving, the accuracy required, the simplifications possible, etc.
• A generic optimization problem model would be in the form given by:

• where x is the decision variable, C(x) is the objective function and g(x) ≤ b is the
inequality constraint.

Engr. Raheel Muzzammel, University of Lahore

Problem Description
• The decision variables may be either real or integer.
• For instance,
✓In an optimum generation scheduling problem, the active power generations are real
✓In a capacitor allocation problem, the number of capacitor banks to be installed in a
specific bus is integer.
• C and g may be either continuous or discrete functions of the decision
variable in an explicit or implicit form; linear or nonlinear.
• Based on those, the optimization problem is appropriately named.
• For instance:
✓An integer linear optimization problem is a problem in which both C and g are linear
functions of integer decision variables.

Engr. Raheel Muzzammel, University of Lahore

Problem Description

Engr. Raheel Muzzammel, University of Lahore

Mathematical Algorithms in Power System
Planning

Engr. Raheel Muzzammel, University of Lahore

Mathematical Algorithms
• A mathematical optimization technique formulates the problem in a
mathematical representation:
➢ Provided the objective function and/or the constraints are nonlinear, the resulting problem is
designated as Non Linear optimization Problem (NLP).
➢ A special case of NLP is quadratic programming in which the objective function is a quadratic
function of x.
➢ If both the objective functions and the constraints are linear functions of x, the problem is
designated as a Linear Programming (LP) problem.
➢ Other categories may also be identified based on the nature of the variables.
➢ For instance, if x is of integer type, the problem is denoted by Integer Programming (IP).
➢ Mixed types such as Mixed Integer Linear Programming (MILP) may also exist in which while
the variables may be both real and integer, the problem is also of LP type.

Engr. Raheel Muzzammel, University of Lahore

Mathematical Algorithms
• Calculus Methods • If inequality constraints are also applicable,
• These types of methods are the traditional way of still the basic method may be used; however,
seeking optimum points. the so called Kuhn-Tucker conditions should
be observed. The solution is not so
• These are applicable to continuous and
differentiable functions of both objective and straightforward in that case.
constraints terms.
• They make use of differential calculus in locating
the optimum points.
• Based on the basic differential calculus developed
for finding the optimum points of C(x), the method
of Lagrange Multipliers has been developed in
finding the optimum points; where equality
constraints may also apply.

Engr. Raheel Muzzammel, University of Lahore

Mathematical Algorithms
• Brief Knowledge about Lagrange Multiplier and Kuhn Tucker Conditions
• In mathematical optimization, the method of Lagrange multipliers is a
strategy for finding the local maxima and minima of a function subject to
equality constraints (i.e., subject to the condition that one or more
equations have to be satisfied exactly by the chosen values of the variables)
• In mathematical optimization, the Karush–Kuhn–Tucker (KKT) conditions,
also known as the Kuhn–Tucker conditions, are first derivative tests
(sometimes called first-order necessary conditions) for a solution in
nonlinear programming to be optimal, provided that some regularity
conditions are satisfied.

Engr. Raheel Muzzammel, University of Lahore

Mathematical Algorithms
• Linear Programming (LP) Method
• LP is an optimization method in which both the objective function
and the constraints are linear functions of the decision variables.
• Any LP problem can be stated as a minimization problem; due to the
fact that, as already described, maximizing C(x) is equivalent to
minimizing (-C(x)).
• All constraints may be stated as equality type; due to the fact that any
inequality constraint of the form given by:

Engr. Raheel Muzzammel, University of Lahore

Mathematical Algorithms
• These inequalities can be transformed to equality constraints, given
by:

Engr. Raheel Muzzammel, University of Lahore

Mathematical Algorithms

• The problem can be stated in a form known as canonical. Then, a solution known
as the simplex method, first devised in 1940s, may be used to solve the problem.
• Using the simplex method normally requires a large amount of computer storage
and time. The so called revised simplex method is a revised method in which
less computational time and storage space are required.
Engr. Raheel Muzzammel, University of Lahore
Mathematical Algorithms
• Canonical Form

• In canonical form, all the constraints are equalities, whereas in

standard form, all the constraints are inequalities.

Engr. Raheel Muzzammel, University of Lahore

Mathematical Algorithms

• The revised simplex method is mathematically equivalent to the

standard simplex method but differs in implementation.
• Instead of maintaining a tableau which explicitly represents the
constraints adjusted to a set of basic variables, it maintains a
representation of a basis of the matrix representing the constraints.
• A set B is a basis if its elements are linearly independent and every
element of Vector Space V is a linear combination of elements of B.

Engr. Raheel Muzzammel, University of Lahore

 Mathematical Algorithms

Basis of Matrix
A basis of a vector space is a set of vectors that are linearly independent and span the
vector space.
Vectors are linearly independent when they cannot be formed using a linear combination
of any of the others.
The span of a set of vectors is the set containing all possible linear combinations of those
vectors.

A space comprised of vectors, collectively with the associative and commutative law of addition of vectors
and also the associative and distributive process of multiplication of vectors by scalars is called vector space.

Engr. Raheel Muzzammel, University of Lahore

Mathematical Algorithms
• Duality Principle
• In linear programming, duality implies that each linear programming
problem can be analyzed in two different ways but would have
equivalent solutions.
• Any LP problem (either maximization and minimization) can be stated in
another equivalent form based on the same data.
• Decomposition Principle
• If the LP problem has a special structure, a so called decomposition
principle may be employed to solve the problem in which less computer
storage is required.
• In this way, the problem can be solved more efficiently.

Engr. Raheel Muzzammel, University of Lahore

Mathematical Algorithms
• Non Linear Programming (NLP) Method
• If the objective function and/or the constraints are nonlinear functions of the
decision variables, the resulting optimization problem is called NLP.
• The solution methods for unconstrained problems may be generally classified as
direct search (or non-gradient) methods and descent (or gradient) methods.
• The former methods do not use the partial derivatives of the objective function
and are suitable for simple problems involving a relatively small number of
variables.
• The latter methods require the evaluations of the first and possibly, the higher
order derivatives of the objective function.
• As a result, these methods are generally more efficient than the direct methods.

Engr. Raheel Muzzammel, University of Lahore

Mathematical Algorithms
• All the unconstrained optimization methods are iterative in nature
and start from an initial trial solution; moving stepwise in a sequential
manner towards the optimum solution.
• The gradient methods have received more attention in power system
literature.
• For instance, in the so called steepest descent method; widely used in
power system literature,
• the gradient vector is used to calculate the optimum step length along the
search direction so that the algorithm efficiency is maximized.

Engr. Raheel Muzzammel, University of Lahore

Mathematical Algorithms
• In constrained optimization, two types of methods, namely, direct and
indirect methods apply.
• In the former methods, the constraints are handled in an explicit manner.
• In most of the latter methods; the constrained problem is converted into a
sequence of unconstrained problems and solved through available
algorithms.
• As an example of the direct methods, in the so called constraint
approximation method, the objective function and the constraints are
linearized about some point.
• The resulting approximated LP problem is solved using LP techniques.
• The resulting solution is then used to construct a new LP problem. The
process is continued until a convergence criterion is satisfied.

Engr. Raheel Muzzammel, University of Lahore

Mathematical Algorithms
• As an example of the indirect methods, the so called penalty function
method, works on the principle of converting the problem into an
unconstrained type.
• It is, in turn, classified as interior and exterior penalty function
methods.
• In the former, the sequence of unconstrained minima lie in the
feasible region
• In the latter, the sequence of unconstrained minima lie in the
infeasible region.
• In both, they move towards the desired solution.

Engr. Raheel Muzzammel, University of Lahore

Mathematical Algorithms
• Dynamic Programming (DP) Method
• Dynamic Programming is a widely used technique in power system studies.
• It is, in fact, a mathematical technique used for multistage decision
problems; originally developed in 1950s.
• A multistage decision problem is a problem in which optimal decisions
have to be made over some stages.
• The stages may be different times, different spaces, different levels, etc.
• The important point is that the output of each stage is the input to the next
serial stage.

Engr. Raheel Muzzammel, University of Lahore

Mathematical Algorithms
• The overall objective function is to be optimized over all stages.
• It is normally a function of the decision variables (x_i) of all stages.
• The important fact is that one can not start from optimizing the first stage;
moving forward toward the final stage; as there may be some correlations
between the stages, too.
• Example
• Let us express a power system example. Suppose we are going to minimize the
generation cost of a power system over a 24-h period.
• Some information is as follows:
✓ There are four generation units available; each of which may be either off or on (so that
various combinations are possible, such as, 1111, 1101, 1001, 0011,…).
✓ The unit efficiencies are different; so that if the system load is low and say, two units can
meet the load, we should use the higher efficient units to supply the load.
✓ The load varies throughout the 24-h period; changing at each hour (stage).

Engr. Raheel Muzzammel, University of Lahore

Mathematical Algorithms
• The multistage decision problem is, in fact, deciding on the units to be on at each
stage so that the overall generation cost over the 24-h period is minimized.
• We note that if no other constraint was imposed, we should optimize our problem
at each stage and sum it over all stages.
• In other words, 24 single stage optimization problems (LP or NLP) have to be
solved to find the final solution.
• Suppose that the final solution looks like Figure in which the unit combinations are
shown at each stage.

Engr. Raheel Muzzammel, University of Lahore

Mathematical Algorithms
• As shown:
• Unit 1 is on at hours 1 and 2, off at hour 3, and on again at hour 4.
• Now what happens if a constraint is imposed expressing the fact that
if unit 1 is turned off, it can not be turned on unless a 5-h period is
elapsed (This type of constraint is called minimum down time of a
unit).
• So, our above solution is not practical.
• Now, how can we find the solution?

Engr. Raheel Muzzammel, University of Lahore

Mathematical Algorithms

The combination 0000 is considered infeasible. This stage is subtracted.

The so called, curse of dimensionality (single stage problem is the solution), in DP problems.

Engr. Raheel Muzzammel, University of Lahore

Mathematical Algorithms
• Integer Programming (IP) Method
• In the algorithms discussed so far, each of the decision variables may take any
real value.
• What happens if a decision variable is limited to take only an integer value?
• For instance, if the decision variable is the number of generation units, taking a
real value is meaningless.
• The optimization algorithms developed for this class of problems are classified as
IP methods.
• If all decision variables are of integer type, the problem is addressed as IP
problem.
• If some decision variables are of integer type while some others are of non-
integer type, the problem is known as mixed integer programming problem.

Engr. Raheel Muzzammel, University of Lahore

 Thank You

Engr. Raheel Muzzammel, University of Lahore