Optimization

with Python
• Linear Programming
• Mixed-Integer Linear Programming
• Nonlinear Programming
• Mixed-Integer Nonlinear Programming
• Heuristics (GA and Particle Swarm)
• Constraint Programming
Rafael Silva Pinto
rafaelcxc@gmail.com
 • Installing Python and Packages
• Starting with Python
• Linear Programming (LP)
• Mixed-Integer Linear Programming (MILP)

Outlines • Nonlinear Programming (NLP)

• Mixed-Integer Nonlinear Programming (MINLP)
• Heuristics (GA and Particle Swarm)
• Constraint Programming
• Practical and good examples
 • Search for the optimal decision
• Any problem of planning:
long, medium, short term, operational
• Applied in the decision-making for investments,
operations, route problems, cost reduction...
What is • Ex.: We want to maximize the revenue for
max 𝑥 + 𝑦
the sale of 2 products (x and y), each
optimization product costs 1 dollar. What is the required
daily production?
−𝑥 + 2𝑦 ≤ 8

2𝑥 + 𝑦 ≤ 14

Constraints 2𝑥 − 𝑦 ≤ 10
2y <= x+8 (time of production)
2x + y <= 14 (raw material) 0 ≤ 𝑥 ≤ 10

2x <= y+10 (historical sales) 0 ≤ 𝑦 ≤ 10

x,y <= 10 (maximum daily production)

PROBLEM PROBLEM
RESOLUTION RESULTS
UNDERSTANDING MODELING
 • Search for the optimal decision
• Any problem of planning:
long, medium, short term, operational
• Applied in the decision-making for investments,
operations, route problems, cost reduction...
What is • Ex.: We want to maximize the revenue for
max 𝑥 + 𝑦
the sale of 2 products (x and y), each
optimization product costs 1 dollar. What is the required
daily production?
−𝑥 + 2𝑦 ≤ 8

𝟐𝒙 + 𝒚𝒙 ≤ 14

Constraints 2𝑥 − 𝑦 ≤ 10
2y <= x+8 (time of production)
x (2 + y) <= 14 (raw material) 0 ≤ 𝑥 ≤ 10

2x <= y+10 (historical sales) 0 ≤ 𝑦 ≤ 10

x,y <= 10 (maximum daily production)

PROBLEM PROBLEM
RESOLUTION RESULTS
UNDERSTANDING MODELING
 • Search for the optimal decision
• Any problem of planning:
long, medium, short term, operational
• Applied in the decision-making for investments,
operations, route problems, cost reduction...
What is • Ex.: We want to maximize the revenue for
max 𝑥 + 𝑦
the sale of 2 products (x and y), each
optimization product costs 1 dollar. What is the required
daily production?
−𝑥 + 2𝑦 ≤ 8

2𝑥 + 𝑦 ≤ 14

Constraints 2𝑥 − 𝑦 ≤ 10
2y <= x+8 (time of production)
0 ≤ 𝑥 ≤ 10
2x + y <= 14 (raw material)
2x <= y+10 (historical sales) 0 ≤ 𝑦 ≤ 10
x,y <= 10 (maximum daily production)
𝑥 and 𝑦 integers

PROBLEM PROBLEM
RESOLUTION RESULTS
UNDERSTANDING MODELING
INTRODUCTION TO
MATHEMATICAL
MODELING
- What is mathematical modeling
- How do we solve mathematical problems?
- Type of variables
- Objective Function and Constraints
- How to model your problem?
- Some examples
- How to learn more?
 WHAT IS I wish to minimize the total cost of

MATHEMATICAL logistics in a food delivery business

MODELING 𝒎𝒊𝒏 ෍ 𝑪𝒐𝒔𝒕(𝒊)

𝒊

𝑪𝒐𝒔𝒕(𝒊) = …..
Real-world problem → Mathematics

Very specific skill

 HOW DO WE SOLVE PROBLEM

OPTIMIZATION
UNDERSTANDING

PROBLEM

PROBLEMS? MODELING

FRAMEWORK

RESOLUTION

SOLVER
First, understand the problem!
RESULTS
Second, convert your problem to math

Third, solve your problem

Fourth, discuss the results

 TYPE OF 𝒎𝒊𝒏 𝑪𝟏 + 𝑪𝟐

VARIABLES 𝑪𝟏 = 𝟎. 𝟏𝑷𝟐𝟏 + 𝟎.5

𝑪𝟐 = 𝑷𝟐 + 𝟑
𝑷𝟏 + 𝑷𝟐 = 𝑷𝑻

What is variable? - Continuous

- Integer (discrete)
What is a parameter? - Binary
- Others
What is an index?

What is a set?
 OBJECTIVE
FUNCTION AND
CONSTRAINTS
max 𝑥 + 𝑦
How to define if a solution is better than other?
−𝑥 + 2𝑦 ≤ 8
What defines if a solution is feasible? 2𝑥 + 𝑦 ≤ 14
2𝑥 − 𝑦 ≤ 10
𝑥, 𝑦 ≥ 0
HOW TO MODEL PROBLEM
UNDERSTANDING

YOUR PROBLEM? PROBLEM

MODELING

RESOLUTION

RESULTS
 EXAMPLE 1
Mayke wish to define the best investments that he should make
with his money. He has a total of 100,000 USD and the
following options for investment.

A) Low risk fund with historical gains of 5% per year

B) Medium risk fund with historical gains 10% per year
C) High risk fund with historical gains of 12% per year

Mayke wish to control the risk of his investments with

maximum of 10% of his money in the investment with high risk,
20% in the investment with medium risk.

Which is the decision of investments in A, B, and C that

maximize the return of investment for Mayke?
 Variables and indexes

EXAMPLE 1 𝑅𝐴 , 𝑅𝐵 , 𝑅𝐶 → Return of investment A,B, and C

𝐶𝐴 , 𝐶𝐵 , 𝐶𝐶 → Invested capital in fund A, B, and C

Mayke wish to define the best investments that he should make Constraints (rules)
with his money. He has a total of 100,000 USD and the 𝐶𝐴 + 𝐶𝐵 + 𝐶𝐶 = 100,000
following options for investment.
𝑅𝐴 = 0.05𝐶𝐴
A) Low risk fund with historical gains of 5% per year 𝑅𝐵 = 0.10𝐶𝐵
B) Medium risk fund with historical gains 10% per year
C) High risk fund with historical gains of 12% per year 𝑅𝐶 = 0.12𝐶𝐶
0 ≤ 𝐶𝐵 ≤ 0.2 ∗ 100,000
Mayke wish to control the risk of his investments with
maximum of 10% of his money in the investment with high risk, 0 ≤ 𝐶𝐶 ≤ 0.1 ∗ 100,000
20% in the investment with medium risk.
Objective Function
Which is the decision of investments in A, B, and C that
maximize the return of investment for Mayke? max(𝑅𝐴 + 𝑅𝐵 + 𝑅𝐶 )
 EXAMPLE 2
Mayke wish to define the best investments that he should make
with his money. He has a total of 100,000 USD and the
following options for investment.

A) Low risk fund with historical gains of 5% per year

B) Medium risk fund with historical gains 10% per year
C) High risk fund with historical gains of 12% per year
D) Especial fund: 10−6 ∗ 𝐶𝐷 2

Mayke wish to control the risk of his investments with

maximum of 10% of his money in the investment with high risk,
20% in the investment with medium risk, and 30% in the
especial fund.

Which is the decision of investments in A, B, and C that

maximize the return of investment for Mayke?
 Index

EXAMPLE 2 𝐴, 𝐵, 𝐶, 𝐷 → funds

Variables
Mayke wish to define the best investments that he should make 𝐶𝐴 , 𝐶𝐵 , 𝐶𝐶 , 𝐶𝐷 → Invested capital in the fund
with his money. He has a total of 100,000 USD and the
following options for investment. 𝑅𝐴 , 𝑅𝐵 , 𝑅𝐶 , 𝑅𝐷 → Return of investment from the fund

A) Low risk fund with historical gains of 5% per year

B) Medium risk fund with historical gains 10% per year
C) High risk fund with historical gains of 12% per year
D) Especial fund: 10−6 ∗ 𝐶𝐷 2

Mayke wish to control the risk of his investments with

maximum of 10% of his money in the investment with high risk,
20% in the investment with medium risk, and 30% in the
especial fund.

Which is the decision of investments in A, B, and C that

maximize the return of investment for Mayke?
 Objective Function

EXAMPLE 2 max 𝑅𝐴 + 𝑅𝐵 + 𝑅𝐶 + 𝑅𝐷

Mayke wish to define the best investments that he should make

with his money. He has a total of 100,000 USD and the
following options for investment.

A) Low risk fund with historical gains of 5% per year

B) Medium risk fund with historical gains 10% per year
C) High risk fund with historical gains of 12% per year
D) Especial fund: 10−6 ∗ 𝐶𝐷 2

Mayke wish to control the risk of his investments with

maximum of 10% of his money in the investment with high risk,
20% in the investment with medium risk, and 30% in the
especial fund.

Which is the decision of investments in A, B, and C that

maximize the return of investment for Mayke?
 Objective Function

EXAMPLE 2 max 𝑅𝐴 + 𝑅𝐵 + 𝑅𝐶 + 𝑅𝐷

Constraints (rules)
Mayke wish to define the best investments that he should make 𝐶𝐴 + 𝐶𝐵 + 𝐶𝐶 + 𝐶𝐷 = 100,000
with his money. He has a total of 100,000 USD and the
𝑅𝐴 = 0.05𝐶𝐴
following options for investment.
𝑅𝐵 = 0.10𝐶𝐵
A) Low risk fund with historical gains of 5% per year
𝑅𝐶 = 0.12𝐶𝐶
B) Medium risk fund with historical gains 10% per year
C) High risk fund with historical gains of 12% per year 𝑅𝐷 = 10−6 𝐶𝐷 2

D) Especial fund: 10−6 ∗ 𝐶𝐷 2

0 ≤ 𝐶𝐵 ≤ 0.2 ∗ 100,000
Mayke wish to control the risk of his investments with 0 ≤ 𝐶𝐶 ≤ 0.1 ∗ 100,000
maximum of 10% of his money in the investment with high risk,
0 ≤ 𝐶𝐷 ≤ 0.3 ∗ 100,000
20% in the investment with medium risk, and 30% in the
especial fund.

Which is the decision of investments in A, B, and C that

maximize the return of investment for Mayke?
 Objective Function

EXAMPLE 2 max 𝑅𝐴 + 𝑅𝐵 + 𝑅𝐶 + 𝑅𝐷

Same as
Mayke wish to define the best investments that he should make
with his money. He has a total of 100,000 USD and the
max ෍ 𝑅𝑓
following options for investment.
𝑓∈𝐹

A) Low risk fund with historical gains of 5% per year

B) Medium risk fund with historical gains 10% per year Set 𝐹 = {A, B, C, D}
C) High risk fund with historical gains of 12% per year
D) Especial fund: 10−6 ∗ 𝐶𝐷 2

Mayke wish to control the risk of his investments with

maximum of 10% of his money in the investment with high risk,
20% in the investment with medium risk, and 30% in the
especial fund.

Which is the decision of investments in A, B, and C that

maximize the return of investment for Mayke?
 Constraints (rules)

EXAMPLE 2 𝐶𝐴 + 𝐶𝐵 + 𝐶𝐶 + 𝐶𝐷 = 100,000
෍ 𝐶𝑓 = 100,000
𝑓∈𝐹

Mayke wish to define the best investments that he should make 𝑅𝐴 = 0.05𝐶𝐴
with his money. He has a total of 100,000 USD and the 𝑅𝐵 = 0.10𝐶𝐵
following options for investment.
𝑅𝐶 = 0.12𝐶𝐶
A) Low risk fund with historical gains of 5% per year 𝑅𝐷 = 10−6 𝐶𝐷 2
B) Medium risk fund with historical gains 10% per year
C) High risk fund with historical gains of 12% per year 𝑅𝑓 = 𝑟𝑒𝑡𝑢𝑟𝑛 𝐶𝑓 ∀𝑓
D) Especial fund: 10−6 ∗ 𝐶𝐷 2

Mayke wish to control the risk of his investments with 0 ≤ 𝐶𝐵 ≤ 0.2 ∗ 100,000
maximum of 10% of his money in the investment with high risk,
0 ≤ 𝐶𝐶 ≤ 0.1 ∗ 100,000
20% in the investment with medium risk, and 30% in the
especial fund. 0 ≤ 𝐶𝐷 ≤ 0.3 ∗ 100,000

Which is the decision of investments in A, B, and C that 𝐶𝑓𝑚𝑖𝑛 ≤ 𝐶𝑓 ≤ 𝐶𝑓𝑚𝑎𝑥 ∀𝑓

maximize the return of investment for Mayke?
 Objective Function

EXAMPLE 2 max ෍ 𝑅𝑓
𝑓∈𝐹

Constraints (rules)
Mayke wish to define the best investments that he should make ෍ 𝐶𝑓 = 100,000
with his money. He has a total of 100,000 USD and the 𝑓∈𝐹
following options for investment.
𝑅𝑓 = 𝑟𝑒𝑡𝑢𝑟𝑛 𝐶𝑓 ∀𝑓
A) Low risk fund with historical gains of 5% per year
𝐶𝑓𝑚𝑖𝑛 ≤ 𝐶𝑓 ≤ 𝐶𝑓𝑚𝑎𝑥 ∀𝑓
B) Medium risk fund with historical gains 10% per year
C) High risk fund with historical gains of 12% per year
D) Especial fund: 10−6 ∗ 𝐶𝐷 2

Mayke wish to control the risk of his investments with

maximum of 10% of his money in the investment with high risk,
20% in the investment with medium risk, and 30% in the
especial fund.

Which is the decision of investments in A, B, and C that

maximize the return of investment for Mayke?
 EXAMPLE 3
You have a company of shoes with 3 very large machines, and you
wish to minimize the total cost of production.

The total cost of production of each machine is:

A) 𝐶𝐴 = 0.1𝑃𝐴2 + 0.5𝑃𝐴 + 0.1
B) 𝐶𝐵 = 0.3𝑃𝐵 + 0.5
C) 𝐶𝐶 = 0.01𝑃𝐶3

where 𝐶 is the cost of production of 𝑃 products for each machine

In the next month, you have a demand of 10,000 shoes. What is

the number of products that should be assigned to each machine
in order to minimize the total cost?
 Variables

EXAMPLE 3 𝐶𝐴 , 𝐶𝐵 , 𝐶𝐶 → Cost of production of machines A,B,C

𝑃𝐴 , 𝑃𝐵 , 𝑃𝐶 → Number of production produced by each
machine (integer)

You have a company of shoes with 3 very large machines, and you
wish to minimize the total cost of production.

The total cost of production of each machine is:

A) 𝐶𝐴 = 0.1𝑃𝐴2 + 0.5𝑃𝐴 + 0.1
B) 𝐶𝐵 = 0.3𝑃𝐵 + 0.5
C) 𝐶𝐶 = 0.01𝑃𝐶3

where 𝐶 is the cost of production of 𝑃 products for each machine

In the next month, you have a demand of 10,000 shoes. What is

the number of products that should be assigned to each machine
in order to minimize the total cost?
 Variables

EXAMPLE 3 𝐶𝐴 , 𝐶𝐵 , 𝐶𝐶 → Cost of production of machines A,B,C

𝑃𝐴 , 𝑃𝐵 , 𝑃𝐶 → Number of production produced by each
machine (integer)

You have a company of shoes with 3 very large machines, and you
Constraints (rules)
wish to minimize the total cost of production.
𝑃𝐴 + 𝑃𝐵 + 𝑃𝐶 = 10,000
The total cost of production of each machine is:
𝐶𝐴 = 0.1𝑃𝐴2 + 0.5𝑃𝐴 + 0.1
A) 𝐶𝐴 = 0.1𝑃𝐴2 + 0.5𝑃𝐴 + 0.1
B) 𝐶𝐵 = 0.3𝑃𝐵 + 0.5 𝐶𝐵 = 0.3𝑃𝐵 + 0.5
C) 𝐶𝐶 = 0.01𝑃𝐶3
𝐶𝐶 = 0.01𝑃𝐶3
where 𝐶 is the cost of production of 𝑃 products for each machine 𝑃𝐴 , 𝑃𝐵 , 𝑃𝐶 ≥ 0

In the next month, you have a demand of 10,000 shoes. What is

the number of products that should be assigned to each machine Objective Function
in order to minimize the total cost? min 𝐶𝐴 + 𝐶𝐵 + 𝐶𝐶
 Variables

EXAMPLE 3 𝐶𝐴 , 𝐶𝐵 , 𝐶𝐶 → Cost of production of machines A,B,C

𝑃𝐴 , 𝑃𝐵 , 𝑃𝐶 → Number of production produced by each
machine (integer)

You have a company of shoes with 3 very large machines, and you
Constraints (rules)
wish to minimize the total cost of production.
෍ 𝑃𝑚 = 10,000
The total cost of production of each machine is: 𝑚
A) 𝐶𝐴 = 0.1𝑃𝐴2 + 0.5𝑃𝐴 + 0.1
𝐶𝐴 = 0.1𝑃𝐴2 + 0.5𝑃𝐴 + 0.1
B) 𝐶𝐵 = 0.3𝑃𝐵 + 0.5
C) 𝐶𝐶 = 0.01𝑃𝐶3 𝐶𝐵 = 0.3𝑃𝐵 + 0.5
𝐶𝐶 = 0.01𝑃𝐶3
where 𝐶 is the cost of production of 𝑃 products for each machine
𝑃𝐴 , 𝑃𝐵 , 𝑃𝐶 ≥ 0
In the next month, you have a demand of 10,000 shoes. What is
the number of products that should be assigned to each machine
in order to minimize the total cost? Objective Function

min ෍ 𝐶𝑚
𝑚 𝑚 = {𝐴, 𝐵, 𝐶}
 Variables

EXAMPLE 3 𝐶𝐴 , 𝐶𝐵 , 𝐶𝐶 → Cost of production of machines A,B,C

𝑃𝐴 , 𝑃𝐵 , 𝑃𝐶 → Number of production produced by each
machine (integer)

You have a company of shoes with 3 very large machines, and you
Constraints (rules)
wish to minimize the total cost of production.
෍ 𝑃𝑚 = 10,000
The total cost of production of each machine is: 𝑚
A) 𝐶𝐴 = 0.1𝑃𝐴2 + 0.5𝑃𝐴 + 0.1
𝐶𝐴 = 0.1𝑃𝐴2 + 0.5𝑃𝐴 + 0.1
B) 𝐶𝐵 = 0.3𝑃𝐵 + 0.5
C) 𝐶𝐶 = 0.01𝑃𝐶3 𝐶𝐵 = 0.3𝑃𝐵 + 0.5
𝐶𝐶 = 0.01𝑃𝐶3
where 𝐶 is the cost of production of 𝑃 products for each machine
𝑃𝐴 , 𝑃𝐵 , 𝑃𝐶 ≥ 0
In the next month, you have a demand of 10,000 shoes. What is
the number of products that should be assigned to each machine
in order to minimize the total cost? Objective Function

min ෍ 𝐶𝑚
𝑚 𝑚 = {𝐴, 𝐵, 𝐶}
 Objective Function

EXAMPLE 3 min ෍ 𝐶𝑚
𝑚

Constraints (rules)
You have a company of shoes with 3 very large machines, and you
wish to minimize the total cost of production. ෍ 𝑃𝑚 = 10,000
𝑚

The total cost of production of each machine is: 𝐶𝐴 = 0.1𝑃𝐴2 + 0.5𝑃𝐴 + 𝛽𝐴 0.1
A) 𝐶𝐴 = 0.1𝑃𝐴2 + 0.5𝑃𝐴 + 0.1 𝐶𝐵 = 0.3𝑃𝐵 + 𝛽𝐵 0.5
B) 𝐶𝐵 = 0.3𝑃𝐵 + 0.5
C) 𝐶𝐶 = 0.01𝑃𝐶3 𝐶𝐶 = 0.01𝑃𝐶3
𝑃𝐴 ≤ 𝛽𝐴 𝑀
where 𝐶 is the cost of production of 𝑃 products for each machine
𝑃𝐵 ≤ 𝛽𝐵 𝑀
In the next month, you have a demand of 10,000 shoes. What is 𝑃𝐴 , 𝑃𝐵 , 𝑃𝐶 ≥ 0
the number of products that should be assigned to each machine
in order to minimize the total cost?

𝑀: very large number 𝑚 = {𝐴, 𝐵, 𝐶}

 EXAMPLE 4
Create a generic formulation to minimize the path from point A to B

2 P1
30
B
A 10 10
5
7 P3
P2 8

The numbers are the distances from one point to another

 Variables

EXAMPLE 4 𝑥𝑖,𝑗 → Binary decision on connection point i to j

Create a generic formulation to minimize the path from point A to B

2 P1
30
B
A 10 10
5
7 P3
P2 8

The numbers are the distances from one point to another

 Variables

EXAMPLE 4 𝑥𝑖,𝑗 → Binary decision on connection point i to j

Parameters
𝐷𝑖,𝑗 → Distance from point i to j
Create a generic formulation to minimize the path from point A to B

2 P1
30
B
A 10 10
5
7 P3
P2 8

The numbers are the distances from one point to another

 Variables

EXAMPLE 4 𝑥𝑖,𝑗 → Binary decision on connection point i to j

Parameters
𝐷𝑖,𝑗 → Distance from point i to j
Create a generic formulation to minimize the path from point A to B

Sets
Ω𝑖𝑛
𝑖 → set of nodes that connect to arcs entering node 𝑖
2 P1
30
B
Ω𝑜𝑢𝑡
𝑖 → set of nodes that connect to arcs exiting node 𝑖
A 10 10
5 Example:
7 P3
Ω𝑖𝑛
𝑃1 = {𝐴, 𝑃2} Ω𝑖𝑛
𝐵 = {𝑃1, 𝑃3}
P2 8 Ω𝑜𝑢𝑡
𝑃1 = {𝑃2, 𝐵} Ω𝐴𝑜𝑢𝑡 = {𝑃1, 𝑃2}

The numbers are the distances from one point to another

 Variables, Parameters and sets

EXAMPLE 4 𝑥𝑖,𝑗 → Binary decision on connection point i to j

𝐷𝑖,𝑗 → Distance from point i to j
Ω𝑖𝑛
𝑖 → set of nodes that connect to arcs entering node 𝑖

Ω𝑜𝑢𝑡
𝑖 → set of nodes that connect to arcs exiting node 𝑖
Create a generic formulation to minimize the path from point A to B

Objective Function

2 P1 B min ෍ 𝑥𝑖,𝑗 𝐷𝑖,𝑗

30 (𝑖,𝑗)

A 10 10 Constraints
5
7 P3 ෍ 𝑥𝐴,𝑗 = 1
P2 8 𝑗∈Ω𝑜𝑢𝑡
𝐴

෍ 𝑥𝑖,𝐵 = 1
𝑖∈Ω𝑖𝑛
𝐵
The numbers are the distances from one point to another
෍ 𝑥𝑖,𝑗 = ෍ 𝑥𝑗,𝑖 ∀𝑖\{𝐴, 𝐵}
𝑗∈Ω𝑜𝑢𝑡
𝑖 𝑗∈Ω𝑖𝑛
𝑖
 EXAMPLE 5
Petter has a construction company. He needs to assign 5 of the
company’s teams to work in some of the constructions below:

A) Revenue of 500, requires 1 team

B) Revenue of 4,000, requires 3 teams
C) Revenue of 3,000, requires 2 teams
D) Revenue of 2,000, requires 1 team
E) Revenue of 2,000, requires 5 teams

Select the constructions that would maximize the revenue.

- Each construction can be made just once
- Not all constructions will be selected
 Variables

EXAMPLE 5 𝑥𝑜 → Binary decision on selecting (or not) construction 𝑜

Petter has a construction company. He needs to assign 5 of the

company’s teams to work in some of the constructions below:

A) Revenue of 500, requires 1 team

B) Revenue of 4,000, requires 3 teams
C) Revenue of 3,000, requires 2 teams
D) Revenue of 2,000, requires 1 team
E) Revenue of 2,000, requires 5 teams

Select the constructions that would maximize the revenue.

- Each construction can be made just once
- Not all constructions will be selected
 Variables

EXAMPLE 5 𝑥𝑜 → Binary decision on selecting (or not) construction 𝑜

Parameters
𝑅𝑜 → Revenue of construction 𝑜
Petter has a construction company. He needs to assign 5 of the 𝑁𝑇𝑜 → Number of teams required for the construction 𝑜
company’s teams to work in some of the constructions below:
𝑁𝑇 𝑚𝑎𝑥 → Total number of teams (available) (5)
A) Revenue of 500, requires 1 team
B) Revenue of 4,000, requires 3 teams
C) Revenue of 3,000, requires 2 teams
D) Revenue of 2,000, requires 1 team
E) Revenue of 2,000, requires 5 teams

Select the constructions that would maximize the revenue.

- Each construction can be made just once
- Not all constructions will be selected
 Variables

EXAMPLE 5 𝑥𝑜 → Binary decision on selecting (or not) construction 𝑜

Parameters
𝑅𝑜 → Revenue of construction 𝑜
Petter has a construction company. He needs to assign 5 of the 𝑁𝑇𝑜 → Number of teams required for the construction 𝑜
company’s teams to work in some of the constructions below:
𝑁𝑇 𝑚𝑎𝑥 → Total number of teams (available) (5)
A) Revenue of 500, requires 1 team
B) Revenue of 4,000, requires 3 teams Objective Function
C) Revenue of 3,000, requires 2 teams
D) Revenue of 2,000, requires 1 team max ෍ 𝑥𝑜 𝑅𝑜
E) Revenue of 2,000, requires 5 teams 𝑜

Constraints
Select the constructions that would maximize the revenue.
- Each construction can be made just once
෍ 𝑥𝑜 𝑁𝑇𝑜 ≤ 𝑁𝑇 𝑚𝑎𝑥
- Not all constructions will be selected
𝑜
 Variables

EXAMPLE 5 𝑥𝑜 → Binary decision on selecting (or not) construction 𝑜

Parameters
𝑅𝑜 → Revenue of construction 𝑜
Petter has a construction company. He needs to assign 5 of the 𝑁𝑇𝑜 → Number of teams required for the construction 𝑜
company’s teams to work in some of the constructions below:
𝑁𝑇 𝑚𝑎𝑥 → Total number of teams (available) (5)
A) Revenue of 500, requires 1 team
B) Revenue of 4,000, requires 3 teams Objective Function
C) Revenue of 3,000, requires 2 teams
D) Revenue of 2,000, requires 1 team max ෍ 𝑥𝑜 𝑅𝑜
E) Revenue of 2,000, requires 5 teams 𝑜

Constraints
Select the constructions that would maximize the revenue.
- Each construction can be made just once
෍ 𝑥𝑜 𝑁𝑇𝑜 ≤ 𝑁𝑇 𝑚𝑎𝑥
- Not all constructions will be selected
𝑜

Solution: Constructions B and C

 Variables

EXAMPLE 6 𝑥𝑜 → Binary decision on selecting (or not) construction 𝑜

Parameters
𝑅𝑜 → Revenue of construction 𝑜
Petter has a construction company. He needs to assign 5 of the 𝑁𝑇𝑜 → Number of teams required for the construction 𝑜
company’s teams to work in some of the constructions below:
𝑁𝑇 𝑚𝑎𝑥 → Total number of teams (available) (5)
A) Revenue of 500, requires 1 team
B) Revenue of 4,000, requires 3 teams Objective Function
C) Revenue of 3,000, requires 2 teams
D) Revenue of 2,000, requires 1 team max ෍ 𝑥𝑜 𝑅𝑜
E) Revenue of 2,000, requires 5 teams 𝑜

Constraints
Select the constructions that would maximize the revenue.
- Each construction can be made just once
෍ 𝑥𝑜 𝑁𝑇𝑜 ≤ 𝑁𝑇 𝑚𝑎𝑥
- Not all constructions will be selected
𝑜
- Construction C can only be selected if A is selected
- Construction D can only be selected if A and C are selected
 Variables
𝑥𝑜 → Binary decision on selecting (or not) construction 𝑜

EXAMPLE 6 Parameters
𝑅𝑜 → Revenue of construction 𝑜
𝑁𝑇𝑜 → Number of teams required for the construction 𝑜
Petter has a construction company. He needs to assign 5 of the
company’s teams to work in some of the constructions below: 𝑁𝑇 𝑚𝑎𝑥 → Total number of teams (available) (5)

A) Revenue of 500, requires 1 team

Objective Function
B) Revenue of 4,000, requires 3 teams
C) Revenue of 3,000, requires 2 teams max ෍ 𝑥𝑜 𝑅𝑜
D) Revenue of 2,000, requires 1 team 𝑜

E) Revenue of 2,000, requires 5 teams

Constraint

Select the constructions that would maximize the revenue. ෍ 𝑥𝑜 𝑁𝑇𝑜 ≤ 𝑁𝑇 𝑚𝑎𝑥
- Each construction can be made just once 𝑜

- Not all constructions will be selected 𝑥𝐶 ≤ 𝑥𝐴

- Construction C can only be selected if A is selected
- Construction D can only be selected if A and C are selected 𝑥𝐷 ≤ 𝑥𝐴

𝑥𝐷 ≤ 𝑥𝐶
Solution: Constructions A, C, and D
 Variables

EXAMPLE 6 𝑥𝑜 → Binary decision on selecting (or not) construction 𝑜

Parameters
𝑅𝑜 → Revenue of construction 𝑜

Petter has a construction company. He needs to assign 5 of the 𝑁𝑇𝑜 → Number of teams required for the construction 𝑜
company’s teams to work in some of the constructions below: 𝑁𝑇 𝑚𝑎𝑥 → Total number of teams (available)

A) Revenue of 500, requires 1 team

B) Revenue of 4,000, requires 3 teams Objective Function
C) Revenue of 3,000, requires 2 teams max ෍ 𝑥𝑜 𝑅𝑜
D) Revenue of 2,000, requires 1 team 𝑜
E) Revenue of 2,000, requires 5 teams
Constraints
Select the constructions that would maximize the revenue.
- Each construction can be made just once ෍ 𝑥𝑜 𝑁𝑇𝑜 ≤ 𝑁𝑇 𝑚𝑎𝑥
- Not all constructions will be selected 𝑜

- Construction C can only be selected if A is selected 𝑥𝐶 ≤ 𝑥𝐴

- Construction D can only be selected if A and C are selected
𝑥𝐷 ≤ 𝑥𝐴 ∗ 𝑥𝐶 (???)
 EXAMPLE 7
Mark wishes to define the scheduling of costumers that he must
attend in the next 3 days.

The list of jobs (demands) with the duration of job and its profit is
defined below:
A) duration 2h, profit 200 USD
B) duration 3h, profit 500 USD
C) duration 5h, profit 300 USD
D) duration 2h, profit 100 USD
E) duration 6h, profit 1,000 USD
F) duration 4h, profit 300 USD

Mark wants to maximize the profit for the next 3 days working 6h
per day. Which demands he should attend per day?
- Neglect the traveling time
- Each demand only can be attended once
 Variables

EXAMPLE 7 𝑥𝑗,𝑑 → Binary decision on attending (or not) job j in day d

Parameters
𝑃𝑗 → Profit for the job
Mark wishes to define the scheduling of costumers that he must 𝐷𝑗 → Duration of the job in hours
attend in the next 3 days. 𝑇ℎ → Number of hours in a working day (6)

The list of jobs (demands) with the duration of job and its profit is
defined below:
A) duration 2h, profit 200 USD
B) duration 3h, profit 500 USD
C) duration 5h, profit 300 USD
D) duration 2h, profit 100 USD
E) duration 6h, profit 1,000 USD
F) duration 4h, profit 300 USD

Mark wants to maximize the profit for the next 3 days working 6h
per day. Which demands he should attend per day?
- Neglect the traveling time
- Each demand only can be attended once
 Variables

EXAMPLE 7 𝑥𝑗,𝑑 → Binary decision on attending (or not) job j in day d

Parameters
𝑃𝑗 → Profit for the job
Mark wishes to define the scheduling of costumers that he must 𝐷𝑗 → Duration of the job in hours
attend in the next 3 days. 𝑇ℎ → Number of hours in a working day (6)

The list of jobs (demands) with the duration of job and its profit is
defined below: Objective Function
A) duration 2h, profit 200 USD
max ෍ ෍ 𝑥𝑗,𝑑 𝑃𝑗
B) duration 3h, profit 500 USD
𝑗 𝑑
C) duration 5h, profit 300 USD
D) duration 2h, profit 100 USD Constraints
E) duration 6h, profit 1,000 USD
F) duration 4h, profit 300 USD ෍ 𝑥𝑗,𝑑 𝐷𝑗 ≤ 𝑇ℎ ∀𝑑
𝑗
Mark wants to maximize the profit for the next 3 days working 6h
෍ 𝑥𝑗,𝑑 ≤ 1 ∀𝑗
per day. Which demands he should attend per day?
𝑑
- Neglect the traveling time
- Each demand only can be attended once
 Variables

EXAMPLE 7 𝑥𝑗,𝑑 → Binary decision on attending (or not) job j in day d

Parameters
𝑃𝑗 → Profit for the job
Mark wishes to define the scheduling of costumers that he must 𝐷𝑗 → Duration of the job in hours
attend in the next 3 days. 𝑇ℎ → Number of hours in a working day (6)

The list of jobs (demands) with the duration of job and its profit is
defined below: Objective Function
A) duration 2h, profit 200 USD
max ෍ ෍ 𝑥𝑗,𝑑 𝑃𝑗
B) duration 3h, profit 500 USD
𝑗 𝑑
C) duration 5h, profit 300 USD
D) duration 2h, profit 100 USD Constraints
E) duration 6h, profit 1,000 USD
F) duration 4h, profit 300 USD ෍ 𝑥𝑗,𝑑 𝐷𝑗 ≤ 𝑇ℎ ∀𝑑
Solution: Profit Total = 2100.0
Mark wants to maximize the profit for the next 3 days working 6h
𝑗

Job E in day 1 (duration 6, profit 1000) ෍ 𝑥𝑗,𝑑 ≤ 1 ∀𝑗

per
Job day. Which
B in day demands
2 (duration he should
3, profit 500) attend per day? 𝑑
-Job
Neglect the2 (duration
D in day traveling2,time
profit 100)
-Job
Each
A indemand only can
day 3 (duration be attended
2, profit 200) once
Job F in day 3 (duration 4, profit 300)
 Variables

EXAMPLE 8 𝑥𝑗,𝑑 → Binary decision on attending (or not) job j in day d

Parameters
𝑃𝑗 → Profit for the job
Mark wishes to define the scheduling of costumers that he must 𝐷𝑗 → Duration of the job in hours
attend in the next 3 days. 𝑇ℎ → Number of hours in a working day (6)

The list of jobs (demands) with the duration of job and its profit is
defined below: Objective Function
A) duration 2h, profit 200 USD
max ෍ ෍ 𝑥𝑗,𝑑 𝑃𝑗
B) duration 3h, profit 500 USD
𝑗 𝑑
C) duration 5h, profit 300 USD
D) duration 2h, profit 100 USD Constraints
E) duration 6h, profit 1,000 USD
F) duration 4h, profit 300 USD ෍ 𝑥𝑗,𝑑 𝐷𝑗 ≤ 𝑇ℎ ∀𝑑
𝑗
Mark wants to maximize the profit for the next 3 days working 6h
per day. Which demands he should attend per day? ෍ 𝑥𝑗,𝑑 ≤ 1 ∀𝑗
- Neglect the traveling time 𝑑
- Each demand only can be attended once
- Mark wish to do a maximum of 1 job per day
 Variables
𝑥𝑗,𝑑 → Binary decision on attending (or not) job j in day d

EXAMPLE 8 Parameters
𝑃𝑗 → Profit for the job
𝐷𝑗 → Duration of the job in hours
Mark wishes to define the scheduling of costumers that he must
attend in the next 3 days. 𝑇ℎ → Number of hours in a working day (6)

The list of jobs (demands) with the duration of job and its profit is
Objective Function
defined below:
A) duration 2h, profit 200 USD max ෍ ෍ 𝑥𝑗,𝑑 𝑃𝑗
B) duration 3h, profit 500 USD 𝑗 𝑑
C) duration 5h, profit 300 USD
D) duration 2h, profit 100 USD Constraints
E) duration 6h, profit 1,000 USD
෍ 𝑥𝑗,𝑑 𝐷𝑗 ≤ 𝑇ℎ ∀𝑑
F) duration 4h, profit 300 USD 𝑗

Mark wants to maximize the profit for the next 3 days working 6h ෍ 𝑥𝑗,𝑑 ≤ 1 ∀𝑗
𝑑
per day. Which demands he should attend per day?
෍ 𝑥𝑗,𝑑 ≤ 1 ∀𝑑
- Neglect the traveling time
𝑗
- Each demand only can be attended once
- Mark wish to do a maximum of 1 job per day
 HOW TO
LEARN MORE?

Try to solve exercises from books

Try to read articles and REALLY understand the proposed models

PRACTICE!!!
Installing Python
and Packages
 WinPython
Portable
How to https://winpython.github.io/

start with Anaconda

Python distribution platform
https://www.anaconda.com/

Python Python Installation

https://www.python.org/
 • Windows:

Installing • Go to https://www.python.org/
• Download and install
• Check version!

Python • Update PIP

• Linux
• Check your version:
python --version
 • Command Prompt:
pip install PACKAGE_NAME

Packages
IDE Spyder pip install spyder
 Jupyter pip install jupyterlab
Notebook
Starting with Python
Starting with Lists
Python Tuples
Dictionaries
Starting with If
Python For
While

Inline commands
Starting with Functions
Python
Starting with Numpy
Python
Starting with Pandas
Python
Starting with Pandas
Python Reading from Excel
and some functions
Starting with Matplotlib
Python
Linear Programming (LP)
Introduction

max 𝑥 + 𝑦
−𝑥 + 2𝑦 ≤ 8

2𝑥 + 𝑦 ≤ 14

2𝑥 − 𝑦 ≤ 10

0 ≤ 𝑥 ≤ 10

0 ≤ 𝑦 ≤ 10
 Linear Solver
vs
Programming Framework
 PROBLEM
UNDERSTANDING

PROBLEM
MODELING

Solver PROGRAMMING
FRAMEWORK
vs LANGUAGE

Framework PROBLEM
SOLVE SOLVER

RESULTS
 Linear Or-Tools
Programming https://developers.google.com/optimization
 Linear Scip
Programming https://www.scipopt.org/

Download and instal SCIP

Install package PYSCIPOPT

Set environment variable SCIPOPTDIR

Package documentation
https://github.com/SCIP-Interfaces/PySCIPOpt
 Gurobi
https://www.gurobi.com/
Download and install Gurobi
Activate Gurobi

Linear CPLEX
https://www.ibm.com/products/ilog-

Programming cplex-optimization-studio
Download and install

GLPK
http://sourceforge.net/projects/winglpk/
Download
Add GLPK/win64 to the
environment variable PATH
 Linear Pyomo

Programming pip install pyomo

Documentation
https://pyomo.readthedocs.io/
 Linear PuLP

Programming pip install cython

pip install pulp

More information in
https://github.com/coin-or/pulp
 Linear Which solver
should I choose?
Programming
Which framework and solver
How easy to How easy to configure a new
Framework (AML) Linear Problems Nonlinear Problems
start with solver and about documentation
Pyomo X X High High
Ortools X Very High Low
PuLP X High High
SCIP X X Very High Not possible / Low
SciPy X X Low Medium

Solver Linear Problemas Nonlinear Free / Commercial

Gurobi X COMMERCIAL
Cplex X COMMERCIAL
CBC X FREE
GLPK X FREE
IPOPT X FREE
SCIP X X FREE
Baron X COMMERCIAL
Exercise
Show the optimal solution and processing time for the following problem:

min −4𝑥 − 2𝑦
Try to solve it by your self
𝑥+𝑦 ≤8
8𝑥 + 3𝑦 ≥ −24 Check the solution in the
−6𝑥 + 8𝑦 ≤ 48 resource of this class
3𝑥 + 5𝑦 ≤ 15
𝑥≤3 Estimated time: 30min
𝑦≥0

Observation: Use the package time to compute the processing time

Working with Pyomo
 Using different
Working with solvers

Pyomo CBC:

https://projects.coin-or.org/Cbc

https://bintray.com/coin-or/download/Cbc/
Working with Arrays and
Pyomo Summations
Arrays and Summation
Power Generation (Pg) 4

min ෍ 𝐶𝑔 𝑖𝑔 𝑃𝑔 𝑖𝑔
ID Cost Power Generation
𝑖𝑔 =0
0 0,10 20 kW
4 2
1 0,05 10 kW
෍ 𝑃𝑔 (𝑖𝑔 ) = ෍ 𝑃𝑑 (𝑖𝑑 )
2 0,30 40 kW
𝑖𝑔 =0 𝑖𝑐 =0
3 0,40 50 kW
4 0,01 5 kW 𝑃𝑑 0 ≤ 𝑃𝑔 0 + 𝑃𝑔 3

Load Points (Pd)

𝑃𝑔 𝑖𝑔 ≥ 0 ∀𝑖𝑔
ID Load Demand *Only generators 0
𝐿𝐼𝑀
0 50 kW and 3 can provide 𝑃𝑔 𝑖𝑔 ≤ 𝑃𝑔 𝑖𝑔 ∀𝑖𝑔
power to load point 0
1 20 kW
2 30 kW
Working with Print your model,
constraints, and
Pyomo summary
Working with Pyomo
constraint rules
Pyomo
 Some NonLinear
Working with equations

Pyomo - binary * continuos

b*x

- continuos²
x ** 2
Mixed-Integer Linear
Programming (MILP)
Introduction
max 𝑥 + 𝑦
−𝑥 + 2𝑦 ≤ 8

2𝑥 + 𝑦 ≤ 14

2𝑥 − 𝑦 ≤ 10

0 ≤ 𝑥 ≤ 10

0 ≤ 𝑦 ≤ 10
Introduction
max 𝑥 + 𝑦
−𝑥 + 2𝑦 ≤ 𝟕
7
(4.2, 5.6) 2𝑥 + 𝑦 ≤ 14

2𝑥 − 𝑦 ≤ 10

0 ≤ 𝑥 ≤ 10

0 ≤ 𝑦 ≤ 10
Introduction max 𝑥 + 𝑦
−𝑥 + 2𝑦 ≤ 7

2𝑥 + 𝑦 ≤ 14
7
(4, 5.5) 2𝑥 − 𝑦 ≤ 10
D
0 ≤ 𝑥 ≤ 10
E
C 0 ≤ 𝑦 ≤ 10
A B
𝑥 as integer
 Pyomo
MILP
model.x = pyo.Var(within=Integers)

model.x = pyo.Var(within=Binary)

https://pyomo.readthedocs.io/en/stable/p
yomo_modeling_components/Sets.html#p
redefined-virtual-sets
 Ortools
MILP
Change the Solver (from GLOP to):
CBC
Gurobi
Cplex

x = solver.IntVar(0,10,'x’)
 SCIP
MILP
x = model.addVar('x', vtype=‘INTEGER’)
Exercise
Find the optimal solution for the following problem
5

min ෍ 𝑥𝑖 + 𝑦
𝑖=1
Try to solve it by your self
5
Tip 1
෍ 𝑥𝑖 + 𝑦 ≤ 20 Check the solution in the
From 1 to 5
𝑖=1
Tip 2 next class
𝑥𝑖 + 𝑦 ≥ 15 , ∀𝑖
For all i
5

෍ 𝑖 ∙ 𝑥𝑖 ≥ 10 Estimated time: 1 hour

𝑖=1
𝑥5 + 2𝑦 ≥ 30
𝑥𝑖 , 𝑦 ≥ 0
𝑥𝑖 integer, ∀𝑖
Nonlinear Programming (NLP)
Introduction
max 𝑥 + 𝑦
−𝑥 + 2𝑦 ≤ 8

2𝑥 + 𝑦 ≤ 14

2𝑥 − 𝑦 ≤ 10

0 ≤ 𝑥 ≤ 10

0 ≤ 𝑦 ≤ 10
Introduction max 𝑥 + 𝑥𝑦
−𝑥 + 2𝑦𝒙 ≤ 8

2𝑥 + 𝑦 ≤ 14

2𝑥 − 𝑦 ≤ 10

0 ≤ 𝑥 ≤ 10

0 ≤ 𝑦 ≤ 10
 Pyomo: IPOPT
Search for ipopt binaries
https://www.coin-

NLP or.org/download/binary/Ipopt

Unzip in C:\

Pyomo
opt = SolverFactory(
‘ipopt',
executable='C:\\ipopt\\bin\\ipopt.exe')
NLP SCIP
 Exercise
Find the optimal solution for the following problem

max cos 𝑥 + 1 + cos 𝑥 𝑐𝑜𝑠(𝑦)

−5 ≤ 𝑥 ≤ 5
−5 ≤ 𝑦 ≤ 5

Explore the following options Try to solve it by yourself

model.x = pyo.Var(initialize=N) N can be any number (0)
opt.options[‘tol’] = N N can be any number (1e-6) Check the solution in the next class
Estimated time: 20 min
Mixed-Integer Nonlinear
Programming (MINLP)
Introduction
max 𝑥 + 𝑦
−𝑥 + 2𝑦 ≤ 8

2𝑥 + 𝑦 ≤ 14

2𝑥 − 𝑦 ≤ 10

0 ≤ 𝑥 ≤ 10

0 ≤ 𝑦 ≤ 10
Introduction max 𝑥 + 𝑥𝑦
−𝑥 + 2𝑦𝑥 ≤ 8

2𝑥 + 𝑦 ≤ 14

2𝑥 − 𝑦 ≤ 10

0 ≤ 𝑥 ≤ 10

0 ≤ 𝑦 ≤ 10

𝒙 integer
 Pyomo: Couenne
https://projects.coin-or.org/Couenne

MINLP https://www.coin-
or.org/download/binary/Couenne/

Unzip in C:\

opt = SolverFactory('couenne',
executable='C:\\couenne\\bin\\couenne.exe')
 Decomposition
MINLP Pyomo + MindtPy
opt = SolverFactory('mindtpy’)

opt.solve(model, mip_solver=‘gurobi',
nlp_solver='ipopt')
MINLP SCIP
 Genetic Algorithm

MINLP pip install geneticalgorithm

https://pypi.org/project/genet
icalgorithm/
 Particle Swarm
MINLP pip install pyswarm

https://pythonhosted.org/pyswarm/
 Ortools
https://developers.google.com/opti
mization/cp/integer_opt_cp

CP
Constraint Programming (CP)
 Ortools
https://developers.google.com/opti
mization/cp/integer_opt_cp

CP
Special Cases
 • Linear or NonLinear Models

Introduction • Special NonLinear Models

• Linearizations
 SCOP
Second-Order Cone Programming

Pyomo + Gurobi
SCOP Example
Suppose that you have 3 machines to manufacture shoes, and the cost of each machine is:
𝐶1 = 0.01𝑛12 + 2𝑛1
𝐶2 = 6𝑛2
𝐶3 = 7𝑛3

where 𝐶𝑖 is cost for production of machine i, 𝑛𝑖 is the number of shoes manufactured in machine i

Each machine has a limit of production of 1.000 shoes.

For a total production of 2.100 shoes, how many shoes should each machine made in order to minimize the total cost?

min 𝐶1 + 𝐶2 + 𝐶3

𝑛1 + 𝑛2 + 𝑛3 = 2100
𝐶1 = 0.01𝑛12 + 2𝑛1
𝐶2 = 6𝑛2
𝐶3 = 7𝑛3
0 ≤ 𝑛1 , 𝑛2 , 𝑛3 ≤ 1000
𝑛1 , 𝑛2 , 𝑛3 𝑎𝑠 𝑖𝑛𝑡𝑒𝑔𝑒𝑟𝑠
 https://www.gurobi.com/resource/
non-convex-quadratic-optimization/

NonConvex QP
NonConvex Quadratic Programming
NonConvex QP Example
Suppose that you have 3 machines to manufacture shoes, and the cost of each machine is:
𝐶1 = 0.01𝑛12 + 2𝑛1 𝐶2 = 6𝑛2 𝒏𝟏 𝐶3 = 7𝑛3

where 𝐶𝑖 is cost for production of machine i, 𝑛𝑖 is the number of shoes manufactured in machine i

Each machine has a limit of production of 1.000 shoes.

For a total production of 2.100 shoes, how many shoes should each machine made in order to minimize the total cost?

min 𝐶1 + 𝐶2 + 𝐶3

𝑛1 + 𝑛2 + 𝑛3 = 2100
𝐶1 = 0.01𝑛12 + 2𝑛1
𝑪𝟐 = 𝟔𝒏𝟐 𝒏𝟏
𝐶3 = 7𝑛3
0 ≤ 𝑛1 , 𝑛2 , 𝑛3 ≤ 1000
𝑛1 , 𝑛2 , 𝑛3 𝑎𝑠 𝑖𝑛𝑡𝑒𝑔𝑒𝑟𝑠
 https://developers.google.com
/optimization/routing/vrp

Routing
Problems
OR-TOOLS VRP
Linearization 𝐶 =𝑏∗𝑥
𝑏 𝑖𝑠 𝑏𝑖𝑛𝑎𝑟𝑦

BigM Similar to

Binary * Continuos −𝑏 ∗ 𝑀 ≤ 𝐶 ≤ 𝑏 ∗ 𝑀
− 1−𝑏 ∗𝑀 ≤𝐶−𝑥 ≤ 1−𝑏 ∗𝑀
𝑏 𝑖𝑠 𝑏𝑖𝑛𝑎𝑟𝑦
Linearization 𝐶 =𝑏∗𝑥
𝑏 𝑖𝑠 𝑏𝑖𝑛𝑎𝑟𝑦

BigM Similar to

Binary * Continuos −0 ∗ 𝑀 ≤ 𝐶 ≤ 0 ∗ 𝑀
− 1−0 ∗𝑀 ≤𝐶−𝑥 ≤ 1−0 ∗𝑀
Linearization 𝐶 =𝑏∗𝑥
𝑏 𝑖𝑠 𝑏𝑖𝑛𝑎𝑟𝑦

BigM Similar to

Binary * Continuos 0≤𝐶≤0

−𝑀 ≤ 𝐶 − 𝑥 ≤ 𝑀
Linearization 𝐶 =𝑏∗𝑥
𝑏 𝑖𝑠 𝑏𝑖𝑛𝑎𝑟𝑦

BigM Similar to

Binary * Continuos 𝐶=0

𝑥 can be any value
Linearization 𝐶 =𝑏∗𝑥
𝑏 𝑖𝑠 𝑏𝑖𝑛𝑎𝑟𝑦

BigM Similar to

Binary * Continuos −1 ∗ 𝑀 ≤ 𝐶 ≤ 1 ∗ 𝑀
− 1−1 ∗𝑀 ≤𝐶−𝑥 ≤ 1−1 ∗𝑀
Linearization 𝐶 =𝑏∗𝑥
𝑏 𝑖𝑠 𝑏𝑖𝑛𝑎𝑟𝑦

BigM Similar to

Binary * Continuos −𝑀 ≤ 𝐶 ≤ 𝑀
0≤𝐶−𝑥 ≤0
Linearization 𝐶 =𝑏∗𝑥
𝑏 𝑖𝑠 𝑏𝑖𝑛𝑎𝑟𝑦

BigM Similar to

Binary * Continuos −𝑀 ≤ 𝐶 ≤ 𝑀
𝐶−𝑥 =0
Linearization 𝐶 =𝑏∗𝑥
𝑏 𝑖𝑠 𝑖𝑛𝑡𝑒𝑔𝑒𝑟

BigM Similar to

Binary * Continuos 𝐶 can be any value

𝐶=𝑥
Binary*Continuous Example
Suppose that you have 3 machines to manufacture shoes, and the cost of each machine is:
𝑪𝟏 = 𝟐𝒏𝟏 𝐶2 = 6𝑛2 𝐶3 = 7𝑛3

where 𝐶𝑖 is cost for production of machine i, 𝑛𝑖 is the number of shoes manufactured in machine i

Each machine has a limit of production of 1.000 shoes.

Machine 2 has a start cost of 1000.
For a total production of 2.100 shoes, how many shoes should each machine made in order to minimize the total cost?

min 𝐶1 + 𝐶2 + 𝐶3
𝑛1 + 𝑛2 + 𝑛3 = 2100 b as binary
𝑏=1 represents that machine 2 is ON
𝑪𝟏 = 𝟐𝒏𝟏 𝑏=0 represents that machine 2 is OFF
𝑪𝟐 = 𝒃 ∗ 𝟔𝒏𝟐 + 𝟏𝟎𝟎𝟎
𝒏𝟐 ≤ 𝒃 ∗ 𝟏𝟎𝟎𝟎
𝐶3 = 7𝑛3
0 ≤ 𝑛1 , 𝑛2 , 𝑛3 ≤ 1000 −𝒃 ∗ 𝑴 ≤ 𝑪𝟐 ≤ 𝒃 ∗ 𝑴
− 𝟏 − 𝒃 ∗ 𝑴 ≤ 𝑪𝟐 − 𝟔𝒏𝟐 + 𝟏𝟎𝟎𝟎 ≤ 𝟏 − 𝒃 ∗ 𝑴
𝑛1 , 𝑛2 , 𝑛3 𝑎𝑠 𝑖𝑛𝑡𝑒𝑔𝑒𝑟𝑠
 𝐶 = 𝑏1 ∗ 𝑏2

Linearization 𝑏1 𝑎𝑛𝑑 𝑏2 𝑎𝑟𝑒 𝑏𝑖𝑛𝑎𝑟𝑖𝑒𝑠

Binary * Binary Similar to

𝐶=𝑧
𝑧 ≤ 𝑏1
𝑧 ≤ 𝑏2
𝑧 ≥ 𝑏1 + 𝑏2 − 1
𝑧, 𝑏1 , 𝑏2 𝑎𝑟𝑒 𝑏𝑖𝑛𝑎𝑟𝑖𝑒𝑠
Binary*Binary Example
Suppose that you have 3 machines to manufacture shoes, and the cost of each machine is:
𝑪𝟏 = 𝟐𝒏𝟏 𝐶2 = 6𝑛2 𝐶3 = 7𝑛3

where 𝐶𝑖 is cost for production of machine i, 𝑛𝑖 is the number of shoes manufactured in machine i

Each machine has a limit of production of 1.000 shoes.

Machine 2 can only be ON if Machine 1 is ON
For a total production of 2.100 shoes, how many shoes should each machine made in order to minimize the total cost?

min 𝐶1 + 𝐶2 + 𝐶3
𝑛1 + 𝑛2 + 𝑛3 = 2100
𝑏𝑖 =1 represents that machine i is ON
𝑏𝑖 =0 represents that machine i is OFF
𝑪𝟏 = 𝟐𝒏𝟏
𝒏𝟏 ≤ 𝒃𝟏 ∗ 𝟏𝟎𝟎𝟎
𝐶2 = 6𝑛2 𝒏𝟐 ≤ 𝒛 ∗ 𝟏𝟎𝟎𝟎
𝒏𝟐 ≤ 𝒃𝟏 ∗ 𝒃𝟐 ∗ 𝟏𝟎𝟎𝟎 𝒛 ≤ 𝒃𝟏
𝐶3 = 7𝑛3 𝒛 ≤ 𝒃𝟐
0 ≤ 𝑛1 , 𝑛2 , 𝑛3 ≤ 1000
𝒛 ≥ 𝒃𝟏 + 𝒃𝟐 − 𝟏
𝑛1 , 𝑛2 , 𝑛3 𝑎𝑠 𝑖𝑛𝑡𝑒𝑔𝑒𝑟𝑠
𝒛 𝒂𝒔 𝒃𝒊𝒏𝒂𝒓𝒚
𝑏1 , 𝑏2 as binary
Binary*Binary Example
A good alternative would be using b2<b1, as the following example.
Using this alternative, you do not need to use b2*b1; however, for this class, I will continue with b2*b1 so we can
practice how to work with the multiplication of two binaries variables

min 𝐶1 + 𝐶2 + 𝐶3
𝑛1 + 𝑛2 + 𝑛3 = 2100
𝑪𝟏 = 𝟐𝒏𝟏
𝒏𝟏 ≤ 𝒃𝟏 ∗ 𝟏𝟎𝟎𝟎
𝐶2 = 6𝑛2
𝒏𝟐 ≤ 𝒃𝟐 ∗ 𝟏𝟎𝟎𝟎
𝒃𝟐 ≤ 𝒃𝟏
𝐶3 = 7𝑛3
0 ≤ 𝑛1 , 𝑛2 , 𝑛3 ≤ 1000
𝑛1 , 𝑛2 , 𝑛3 𝑎𝑠 𝑖𝑛𝑡𝑒𝑔𝑒𝑟𝑠
𝑏1 , 𝑏2 as binary
Advanced Features for Pyomo
Case Study
Case Study
Suppose you have 4 machines on your computer’s factory

You have to define the number of computers that each machine must produce during the next 10 hours in order to maximize the total production. The
objective function of our problem is given by
𝑚𝑎𝑥 ෍ ෍ 𝑥𝑚,𝑡
𝑚 𝑡
where 𝑥𝑚,𝑡 represents the number of computers produced by a machine m at hour t.

Some machines have dependencies on others and the constraints are (for each 𝑡)

2𝑥
2𝑥2,𝑡−
2,𝑡 −8𝑥 ≤ 00
8𝑥3,𝑡 ≤ (1)
𝑥2,𝑡 − 2𝑥3,𝑡−2 + 𝑥4,𝑡 ≥ 1 (2)
𝑥2,𝑡 − 2𝑥3,𝑡−2 + 𝑥4,𝑡 ≥ 1 (2)
And the capacity production is given by (for all 𝑡)
෍𝑥𝑥𝑚,𝑡
෍ ≤ 50
𝑚,𝑡 ≤ 50 (3)
𝑚
𝑚

𝑥1,𝑡 + 𝑥2,𝑡−1 + 𝑥3,𝑡 + 𝑥4,𝑡 ≤ 10 (4)

0 ≤ 𝑥𝑚,𝑡 ≤ 10 (5)
Solve the problem and find the optimal production of computers for each machine for each hour of the next 10 hours.

Note the 𝑥2,𝑡−1 , in the last constraint, represents the second machine at hour t-1
Obs.: Constraints (2) and (4) does not exist for t<3 and t<2, respectively
Case Study

𝑚𝑎𝑥 ෍ ෍ 𝑥𝑚,𝑡
𝑚 𝑡

2𝑥2,𝑡 − 8𝑥3,𝑡 ≤ 0 ∀𝑡

𝑥2,𝑡 − 2𝑥3,𝑡−2 + 𝑥4,𝑡 ≥ 1 ∀𝑡 > 2

෍ 𝑥𝑚,𝑡 ≤ 50 ∀𝑡
𝑚

𝑥1,𝑡 + 𝑥2,𝑡−1 + 𝑥3,𝑡 + 𝑥4,𝑡 ≤ 10 ∀𝑡 > 1

0 ≤ 𝑥𝑚,𝑡 ≤ 10 ∀𝑚, ∀𝑡
 Solver
progress
opt.solve(model, tee=True)
Gap Limit opt.options['MIPgap'] = 0.0001

https://www.gurobi.com/documentation/9.1/refman/
mipgap2.html
Time Limit opt.options[‘TimeLimit'] = 60
Inequallity 𝐴≤𝑥≤𝐵
model.C = pyo.Constraint(pyo.inequality(A,x,B))
 ෍ ෍ 𝑥𝑚,𝑡

summation 𝑚 𝑡

pyo.summation(x)
Parameters pyo.Param(initialize=VALUE)

and Sets
pyo.Set(initialize=LIST)

pyo.RangeSet(BEGIN,END)
Constraint’s Rules SAME AS
Rules in the Objective Function

pyo.Objective(expr = pyo.summation(x), sense=pyo.maximize)

SAME AS

pyo.Objective(rule=myobj, sense=pyo.maximize)

def myobj(model):
return pyo.summation(model.x)
 model.C1 = pyo.ConstraintList()
for t in model.setT:
Constraint’s Rules model.C1.add(expr = 2*x[2,t] - 8*x[3,t] <= 0)

with indexes SAME AS

model.C1 = pyo.Constraint(model.setT, rule=myrule)

def myrule(model, t):

return 2*model.x[2,t] - 8*model.x[3,t] <= 0

pyo.Constraint(RangeIndex1, RangeIndex2, ... , rule=myrule)

 https://pyomo.readthedocs.io/en/stable/
working_models.html#warm-starts

Warmstart
Differential https://pyomo.readthedocs.io/en/stable/
modeling_extensions/dae.html

Algebraic
Equations
(DAE)
Practical Examples
Fence in the Garden
What is the largest area that we can fence in a garden using 100 meters of fence?
Define the dimensions of this garden as well.

Note: The garden is already fenced by a wall of rocks in one of its sides.

m𝑎𝑥 𝑥𝑦

2𝑥 + 𝑦 ≤ 100

𝑥, 𝑦 > 0
Maximize Revenue

A car rental company wish to to maximize its revenue.

From the historical sales data, it is known that practicing a rent price (p) between 50 and 200 dollar,
the number of cars rented per day is N(p) = 1001-5p.

What is the rent price that maximizes the daily revenue?

And what is the expected number of cars to be rented?

m𝑎𝑥 𝑝 𝑁
𝑁 = 1001 − 5𝑝
50 ≤ 𝑝 ≤ 200
𝑁 𝑖𝑛𝑡𝑒𝑔𝑒𝑟
Route Optimization

650 m
DISTRICT FOUR 500 m
DISTRICT TWO
DISTRICT SEVEN
220 m 900 m DISRTICT FIVE
DISTRICT ONE
400 m
1500 m

DISTRICT SIX
400 m
DISTRICT THREE

500 m
Route Optimization min ෍ 𝑥𝑖𝑗 𝐷𝑥𝑖𝑗

෍ 𝑥𝑖𝑗 = 1 𝑜𝑟𝑖𝑔𝑖𝑛 𝑛𝑜𝑑𝑒

𝑜𝑢𝑡
Which is the best route from point 1 to point 7?
෍ 𝑥𝑖𝑗 = 1 𝑑𝑒𝑠𝑡𝑖𝑛𝑎𝑡𝑖𝑜𝑛 𝑛𝑜𝑑𝑒
𝑖𝑛

෍ 𝑥𝑖𝑗 ≤ 1 ∀𝑛𝑜𝑑𝑒/(𝑜𝑟𝑖𝑔𝑖𝑛, 𝑑𝑒𝑠𝑡𝑖𝑛𝑎𝑡𝑖𝑜𝑛)

𝑜𝑢𝑡

X57
෍ 𝑥𝑖𝑗 ≤ 1 ∀𝑛𝑜𝑑𝑒/(𝑜𝑟𝑖𝑔𝑖𝑛, 𝑑𝑒𝑠𝑡𝑖𝑛𝑎𝑡𝑖𝑜𝑛)
X36
𝑖𝑛

෍ 𝑥𝑖𝑗 = ෍ 𝑥𝑖𝑗 ∀𝑛𝑜𝑑𝑒/(𝑜𝑟𝑖𝑔𝑖𝑛, 𝑑𝑒𝑠𝑡𝑖𝑛𝑎𝑡𝑖𝑜𝑛)

𝑖𝑛 𝑜𝑢𝑡

𝑎𝑙𝑙 𝑥𝑖𝑗 𝑏𝑖𝑛𝑎𝑟𝑦

Linear Optimal Power Flow: Power Systems
.
What are the optimal values for the generation power units in the following system?
Consider the objective function as the minimization of the power generation cost.

min ෍ 𝐶𝑔 𝑃𝑔
Pg1 <= 20 kW Pg2 <= 30 kW
𝑔
Cg = 0,2 $/kWh Cg = 0,5 $/kWh
~ ~
B0
L0
B1 ෍ 𝑃𝑔 − ෍ 𝑃𝑙 + ෍ 𝑃𝑙 = ෍ 𝑃𝑑 ∀𝑛
Pl = 15 kW 𝑔∈Ω𝐺
𝑛 𝑙∈Ω𝐿𝑛=𝑙 𝑠 𝑙∈Ω𝐿𝑛=𝑙 𝑟 𝑑∈Ω𝐷
𝑛

L1 L2
Pl = 15 kW Pl = 15 kW
𝑃𝑙 = 𝐵𝑙 (𝜃𝑙 𝑛=𝑠 − 𝜃𝑙 𝑛=𝑟 ) ∀𝑙
B2
0 ≤ 𝑃𝑔 ≤ 𝑃𝑔𝑚𝑎𝑥 ∀𝑔
Pd = 25 kW
−𝑃𝑙𝑚𝑎𝑥 ≤ 𝑃𝑙 ≤ 𝑃𝑙𝑚𝑎𝑥 ∀𝑙

Bl = 100 pu −𝜋 ≤ 𝜃𝑛 ≤ 𝜋 ∀𝑛
Sbase = 1 kW
𝜃𝑛 = 0 𝑛: 𝑟𝑒𝑓(0)
 Congratulations!!

Challenges:
https://math.libretexts.org/Courses/Mount_Royal_University/MATH_1200%3A_Calculus_for_Scientists_I/3%3A_Appli
cations_of_Derivatives/3.6%3A_Applied_Optimization_Problems