Excel based Multiobjective Optimization:

Application to a Biodiesel Production from

WCO

Dipesh S Patle
PhD (USM Malaysia)
PDF (POLIMI Italy)
http://www.mnnit.ac.in/index.php/department/engineering/cm/cmfp
 Outline
• Introduction to MOO
1

• MOO methodologies
2

• EMOO
3

• EMOO of a Biodiesel process

4
 Areas of Interest
– Conversion of biomass to biofuels
– Process Optimization
– Process Retrofitting
– Plantwide Control
– Operator Training Simulators
 WORK EXPERIENCE
Post-PhD: 3+ years
 Assistant Professor at MNNIT Allahabad
[Duration: 11/05/2018 to current]

 Associate Professor at VIT University, Vellore.

[Duration: 30/06/2015 to 30/04/2018]

4
Pre-PhD: 4.5 YEARS
 Senior Research Fellow at Indian Institute of Technology (IIT) Kharagpur (BARC
sponsored project entitled ‘Design, Modeling and Control of High Pressure Pilot Scale
HIX Reactive Distillation’)
[Duration: From 06/08/2010 to 31/01/2012]

 Lecturer at Birla Institute of Technology and Science Pilani.

[Duration: From 01/01/2010 to 31/07/2010]

 Teaching Assistant at Birla Institute of Technology and Science Pilani.

[Duration: From 01/01/2008 to 31/12/2009]

 Project Assistant at Birla Institute of Technology and Science Pilani.

[Duration: From 18/08/2007 to 31/12/2007]
 Awards & Achievements
 Received a prestigious Young Scientist Award from Venus International
Foundation on August 11, 2018.

 Received Sanggar Sanjung Award 2014 (Journal Publication 2014) from

Universiti Sains Malaysia.

 Erasmus Mundus Post Doctoral fellowship at Politecnico De Milano,

Italy in 2016

 Entry 'Operator Training Simulator' was selected as one of the finalists

for the Training and Development Award in IChemE Singapore Awards
in 2016.

 Graduation on Time award (GOT) from USM.

 Achieved first prize in an international level paper presentation at

COLLISION 2006 held at DDU Nadiad.
 Research grants
Sr On-going/ Duration Title Agency PI/C Amount
No Completed (years) o-PI (INR)
1 On-going 3 In Situ Biodiesel Production from DST-SERB, PI 9.71 L
Microalgae using Ultrasonic India
Assisted Reactor
2 On-going 2 Direct Biodiesel Synthesis from Research Co- MYR
Microalgae Through In Situ University PI 70,000
Transesterification using (RUI), USM (i.e.
Homogeneous Catalyst Malaysia ~12.5 L)
3 On-going 1 Ultrasound Assisted In Situ Fatty TEQIP III PI 2.00 L
Acid Methyl Ester Production
form Microplytes: A Dry Route
Approach vs a Wet Route
Approach.
4 Completed 1 Direct Biodiesel Synthesis from SEED PI 2.85 L
Microalgae Through In Situ Grant, VIT
Transesterification using Vellore
Homogeneous Catalyst
International collaborations:
 Prof G P Rangaiah from NUS Singapore
 Dr Zainal Ahmad from USM Malaysia
 Prof Davide Manca from POLIMI Italy
 Dr Salman Nazir from Buskerud and Vestfold University College, Norway
 Dr Ashish Gujrathi from SQU Oman
Introduction

What is Process Optimization?

Find the values of process variables that provide the best value
of a performance criterion (e.g., minimize cost or energy,
maximize profit or yield) using quantitative models and
methods.
Example: Design of a Storage Tank
A cylindrical tank for storing 1000 m3 of a liquid, is required. Side
of the tank costs $20/m2, the top $22/m2 and the bottom
$30/m2. Find the optimal tank design.
Introduction
Introduction
Introduction

Why is Optimization Necessary?

Improved Plant Performance
 Better Yield of Valuable Products
 Reduced Energy Consumption
 Longer Time between Shutdowns
 Reduced Maintenance Costs
 Greater Profit
 Etc.
Better Understanding of the Process
 Identify Sensitive Regions and Variables
 Estimating Parameters in Models
 Reconciliation of Measured Data
Introduction

Typical Applications
 Pipeline Sizing and Layout
 Equipment/Plant Design
 Equipment/Plant Operation
 Plant Locations
 Supply of Raw Materials and/or Products
 Minimization of Inventory Costs
 Maintenance and Replacement Scheduling
 Data Validation and Modeling
Introduction

Typical Hierarchy of Control and Optimization

Introduction
Introduction

 Feasible Region:

Sets of Variables Satisfying All Constraints

N > Number of Equality Constraints

 Solution: Unique or Multiple

Introduction

Steps in Process Optimization

 Formulation
 Analysis to Identify Variables & Characteristics
 Select the Objective Function/Performance Criterion
 Develop Process Model & Requirements
 All Constraints
 Check for Degrees of Freedom

Sequential or Iterative?

“The mere formulation of a problem is far more often

essential than its solution, which may be merely a matter of
mathematical or experimental skill.” – Einstein
Introduction

Solution and Review

 Select and Use a Numerical Technique
 Linear Programming
 Nonlinear Programming
 Integer Programming
 Mixed-Integer Linear/Nonlinear Programming
 Global Optimization
 Multi-objective Optimization
 Review/Improve the Solution
 Sensitivity Analysis: Effect of Data used in the
Problem on the Solution
General Problem
General Problem
General
GeneralProblem
Problem
Optimality Criteria
Optimality Criteria
Methods Overview

 Can optimization problems be solved using necessary

and sufficient conditions?

 Numerical methods (techniques) are required for

solving application problems

 Significant research and developments in optimization

methods since 1960’s

 Robust methods and programs are now available for

solving application problems
Methods Overview

Typical Algorithm of a Method

1. Choose an Initial (Trial) Point

2. Iterations to improve trial point

o Based on values of objectives and constraints
o Derivatives may be used in iterations

3. Check for convergence to terminate iterations

 Methods Overview

Remarks
1. Many methods for different types of
problems
2. Each method/technique has certain algorithm
parameters (convergence criterion, maximum
number of iterations, etc.)
3. A method may or may not be successful
 Methods Overview
Characteristics and Classification of Optimization Problems and Methods

Characteristic Classification
Number of variables: one or more Single variable or multivariable optimization
Type of variables: real, integer or mixed Nonlinear, integer or mixed (nonlinear) integer
programming
Nature of equations: liner or nonlinear Linear or nonlinear programming
Constraints: no constraints (besides bounds) Unconstrained or constrained optimization
or with constraints
Number of objectives: one or more Single objective or multi-objective optimization

Derivatives: without or using derivatives Direct or gradient search optimization

Optimum: local or global in the search space Local or global optimization
Random numbers: without or using random Deterministic or stochastic optimization
numbers methods
Trial points/solutions: one or more in Single point (also known as trajectory) or
each iteration population based methods
Methods Overview

 Linear Programming
 Simplex Method

 Single Variable Optimization

 Golden Section Method
 Polynomial Approximation Methods
 Newton Method
Methods Overview

 Unconstrained Optimization
 Direct Search Methods
 Nelder-Mead Simplex Method
 Conjugate Directions Method

 Gradient Search methods

 Steepest Descent Method
 Conjugate Gradient Method
 Quasi-Newton Methods
Methods Overview

 Constrained Optimization
 Direct Search Methods
 Complex Method of Box
 Luus-Jaakola Method
 Stochastic Methods (Metaheuristics)

 Gradient Search methods

 Successive Linear Programming
 Penalty Function Methods
 SQP (Successive Quadratic Programming)
 GRG (Generalized Reduced Gradient)
Methods Overview

 Mixed-Integer Nonlinear Programming

 Branch-and-Bound Methods
 Outer Approximation Method
 Disjunctive Programming

Stochastic Methods (Metaheuristics)

Methods Overview
 Stochastic Methods (Metaheuristics)
Optimization Programs

Solver in MS Excel
 Premium Solver
 Excel-based MOO (EMOO and IMODE)

MATLAB
 GAMS
 Aspen HYSYS and Plus

Programs in FORTRAN, C++, R etc.

 Open-source programs on Internet
Genetic Algorithms

Motivation and ideas for genetic algorithms (GA), as

the name indicates, comes from nature – evolution
of species, survival of the fittest, chromosomes,
genes etc.

Concept of GA was first suggested by Holland (1975)

and it was later developed by his students and other
researchers. GAs have been successfully used for
solving many (chemical) engineering problems,
particularly in the past decade.
Genetic Algorithms

GA works on a population of points (individuals or

chromosomes) and combines (crossover) and modifies
(mutation) some individuals as per specified genetic
operations to generate a new population with better
solutions.

Optimal solutions (fitter individuals) would be found

after many generations. Using schema theorems, GA is
shown to yield better solutions as evolution continues
since the good traits of fitter individuals are passed on
to the next generation.
Genetic Algorithms

Principles of GA
Implementation of evolution process – survival of the
fittest via chromosomes and genes, on a digital
computer for optimization requires new principles.

In spirit with evolution, GA is discussed for maximizing

a function, f(x) subject to bounds on variables only.

For 2 variables, this will be: maximize f(x, y) subject to

xl ≤ x ≤ xu and yl ≤ y ≤ yu.
Genetic Algorithms

Typical Steps in GA
Step 1: Choose schemes for coding, selection, crossover,
mutation etc.

Step 2: Select population size (~ 100), maximum number

of generations (~ 100), string length for coding (as per
accuracy requirements), crossover and mutation
probabilities (pc ~ 0.8 and pm ~ 0.05) etc.

Step 3: Generate initial population randomly and evaluate

fitness function of each individual.
Genetic Algorithms

Step 4: Terminate if the specified number of generations

is exceeded. Else, continue.

Step 5: Increment the generation counter and repeat

from Step 3.

Common Objectives: profit, capital cost, operating cost,

productivity, conversion, yield, selectivity, efficiency, global
warming potential, greenhouse gas emissions etc.
Genetic Algorithms

Generation
a) Create the mating pool. (What is its size?)
b) Select and remove 2 chromosomes from the mating
pool.
c) Perform crossover on the selected chromosomes to
generate off-springs.
d) Apply mutation on each string of the off-spring.
e) Evaluate fitness function of each off-spring.
f) Repeat steps (b) to (e) until mating pool is empty.
g) Off-springs form the new population.
 Sustainable Processes

Three Targets for Sustainable Processes

Future Objectives for Travelling:

Cost, Time, Comfort, CO2 Foot-print etc.
MOO Solutions

Often, a MOO problem has many optimal solutions,

known as non-dominated (or Pareto-optimal) solutions.

Many feasible solutions are inferior to (or dominated by)

others. Only the optimal solutions are better than all others.
 Single vs. Multi-Objective Optimization

Feature Single Objective Optimization Multi-Objective Optimization

(SOO) (MOO)
Number of Only one objective function Two or more objective functions
Objectives
Decision Continuous and/or integer decision variables with bounds on them
Variables
Constraints Equality and/or inequality constraints; some problems may have no
constraints
Type of Optimal Local or global solution Local or global Pareto-optimal front
Solutions
Multi- Only one space for decision Two spaces: one for objective
dimensional variables functions and another for decision
Spaces variables
Techniques Deterministic and stochastic optimization techniques

Development of Developed for solving SOO SOO techniques are extended for
Techniques problems solving MOO problems
 Single vs. Multi-Objective Optimization

Feature Single Objective Optimization Multi-Objective Optimization

(SOO) (MOO)
Computational Deterministic techniques faster Deterministic techniques faster if
Time for Finding compared to stochastic techniques only a few Pareto-optimal solutions
Solutions but the latter more likely to find are required, whereas stochastic
the global solution techniques faster for finding many
Pareto-optimal solutions.
Number of Usually, only one solution Usually, many Pareto-optimal
Solutions solutions
Knowledge on Limited to one optimal solution Quantitative trade-off among many
Optimal Solutions optimal solutions
Selection of a Straightforward since only one or Additional technique/preferences
Solution for a few optimal solutions are found are required to select one of the
Implementation many optimal solutions obtained.

Rangaiah G.P., Sharma S. and Sreepathi B.K., Multi-objective Optimization for the Design and Operation of Energy
Efficient Chemical Processes and Power Generation, Current Opinion in Chemical Engineering, 10, 49-62 (2015).
Multi-Objective Differential Evolution
 MOO Programs
NSGA-II in C (IIT Kanpur)
SPEA2 (ETH Zurich)
gamultiobj in Matlab
NGPM in Matlab
jMetal in Java (University of Malaga, Spain)
NIMBUS via internet (University of Jyvaskyla,
Finland)
EMOO in Excel (NUS Singapore)
IMODE in Excel (NUS Singapore)
EMOO (& IMODE) Programs

 Excel worksheets for user inputs, calculating objective

functions and constraints, MOO results etc.
 VBA/macros for implementing algorithm steps, interfacing
with simulators etc.
 Both programs include two termination criteria based on
search progress (besides maximum number of generations)
 EMOO based on non-dominated sorting genetic algorithm
(NSGA)
 EMOO has both binary and real coding

Sharma S., Rangaiah G.P. and Maréchal F., Multi-Objective Optimization Programs and their Application to Amine
Absorption Process Design for Natural Gas Sweetening, in “Multi-Objective Optimization: Techniques and
Applications in Chemical Engineering” edited by G.P. Rangaiah, Second Edition, World Scientific, 2017.
 EMOO Program
Excel worksheets: calculate objective VBA/macros: implement algorithm steps,
functions and constraints, interfacing, etc. interfacing with simulators, etc.
Framework for EMOO
 Application to
Biodiesel Production from Waste
Cooking Oil
 Motivation

Biodiesel
o Current fuel scarcity
o Non-renewability
o Highly polluting nature
o Fluctuating cost

 WCO does away with food-fuel issue. At last, EU limited the

edible oil’s use for Biodiesel production
 Effective WCO utilization

Process System Engineering Study

o Need for Detailed Simulation
o Need for Multi-objective Optimization
o Need for Plantwide control
o Need for Operator Training Simulator
Process Design and Simulation
o Biodiesel plant capacity is assumed to be 120
kt/annum.

o Dortmund modified UNIFAC is used.

o Vacuum operation for the methanol recovery and

products purification was applied.

o Detailed constituents of WCO is considered.

o Unlike most of the published paper, a detailed kinetics
(10 and 96 reactions to represent esterification and
tranesterification, respectively) is considered.
 MEOH
25 oC

Contd..
1.1 bar
124.78 kg/h
H2SO4 0.2
25 oC P-1 RFFA bar
1.1 bar 2
90 kg/h FRAC-2-1
P-3
OIL FRAC-2
0.236 9
25 oC H-1 FRAC-2-2

Biodiesel process 1.1 bar

15000 kg/h
P-2
RFFA1
bar
W-1-1
0.3 P-5
bar
2
FRAC-1-1
GLY-IN W-1 P-4 H-2
FRAC-1
25 oC 7 0.329
1.1 bar bar
W-1-2
177 kg/h FRAC-1-2 P-6-1
ME-WAT-1 P-6
o
110 C S-1-1
0.8 bar R-CAO-1
55.62 kg/h
S-1 CAO
F-1 R-CAO 30 oC
GLY-2 CASO4 1.1 bar
60 oC 12.48 kg/h
1 bar
30.3 kg/h

BIO-D
49.4 oC
1 bar
R-1 WATER 15167.3 kg/h
RTRANS1 30 oC
1.1 bar
D-1-2 WASTE
D-1-1 180.15 kg/h WASH-2
D-1 (organic)
49.9 oC
1 bar
R-2
197.75 kg/h
RTRANS2 S-4-1
D-2-2
D-2
D-2-1

S-4
0.25 RTRANS3 NA3PO4
bar 50 oC
D-M-1 ME-WAT-2 1 bar
2
122.52 kg/h
FRAC-4 30 oC
H3PO4
0.315 1.1 bar
bar 9 H3PO4-2 73.13 kg/h
30 oC
1.1 bar R-CAT
FRAC-4-2 42.04 kg/h E-62
P-7 R-3
0.34
Patle, D. S., Sharma, S., Ahmad, Z., & P-11
H-5
R-CAT2
bar
2
R-C2-1 FRAC-3-1
Rangaiah, G. P. (2014). Multi-objective H-3 H-4
GLYC-OUT FRAC-3
optimization of two alkali catalyzed 50 Co P-9 10 0.44
bar
1 bar 25 oC
processes for bio-diesel from waste cooking S-3
1637.4 kg/h
P-10
NAOH
1.1 bar
141 kg/h
FRAC-3-2
oil. Energy Conversion and Management, NA3PO4-2
50 oC
MEOH-2
25 oC
1 bar 1.1 bar
Elsevier, 85, 361-372. 80.72 kg/h 1643.73 kg/h
 Process Simulators….

“A model should be as simple as possible and no simpler”

~ Albert Einstein ~

"Essentially, all models are wrong, but some are useful.“

[Box, George E. P.; Norman R. Draper (1987). Empirical Model-Building and Response Surfaces, p.
424, Wiley. ISBN 0471810339]
Multiobjective Optimization

Framework for EMOO

 Objectives, DVs and Constraints

Objective function Decision variables Constraints

Case A: 96 ≤ WCO ≤ 120 kt/annum Mass purity: xBiodiesel ≥ 0.99

Max. Profit (million 55 ≤ TRFFA ≤ 65oC Mass purity: xGlycerol ≥ 0.95

(Methanol Recovery)FRAC-1, FRAC-3, FRAC-3,
USD/annum) 45 ≤ TRTRANS1 ≤ 60oC
FRAC-4 ≥ 0.98
Min. Heat Duty (MW) 45 ≤ TRTRANS2 ≤ 60oC
TFRAC-1 ≤ 150oC
45 ≤ TRTRANS3 ≤ 60oC
TFRAC-2 ≤ 250oC
Case B: 1.5 ≤ (Residence Time)RFFA ≤ 2.5 h TFRAC-3 ≤ 250oC
Max. Profit (million 1.5 ≤ (Residence Time)RTRANS1 ≤ 2.5 h TFRAC-4 ≤ 150oC
USD/annum) 1.5 ≤ (Residence Time)RTRANS2 ≤ 2.5 h
Min. Organic Waste 1.5 ≤ (Residence Time)RTRANS3 ≤ 2.5 h
(tons/annum) 2 ≤ (Feed Stage)FRAC-1 ≤ 7
2 ≤ (Feed Stage)FRAC-2 ≤ 9
2 ≤ (Feed Stage)FRAC-3 ≤ 10
2 ≤ (Feed Stage)FRAC-4 ≤ 9
 Contd.. Component
Start

Module
Addition Selection
Process
Simulation Process
Property
Parameters
Method
and Reaction
Selection
Kinetics
Costing Procedure Results
NO
Validation

YES

Variation of Process
Parameters targeting min. TAC
such as: Number of stages,
Feed stage and Pressure of a
Distillation Column

Optimality NO
(min. TAC)

YES

Cost Estimation
of Optimal
Process Model

End
Contd..
ln 𝐴 +0.09861 ln 𝐴 2
𝐶𝐵 = 𝑒𝑥𝑝 11.0545 −0.9228

𝐶𝑃 = 𝐶𝐵 𝐹𝑃 𝐹𝑀 𝐹𝐿
𝐴
𝐹𝑀 = 𝑎 + 𝑏
100
2
𝑃 𝑃
𝐹𝑃 = 0.9803 + 0.018 + 0.0017
100 100

𝐶𝐵𝑀 = 𝐶𝑝 𝐹𝐵𝑀
𝑎𝑙𝑙 𝑒𝑞𝑢𝑖𝑝𝑚𝑒𝑛𝑡𝑠

𝐶𝑇𝑀 = 1.18𝐶𝐵𝑀

𝐹𝐶𝐼 𝑜𝑟 𝐶𝐺𝑅 = 𝐶𝑇𝑀 + 0.50 𝐶 0 𝐵𝑀

𝑎𝑙𝑙 𝑒𝑞𝑢𝑖𝑝𝑚𝑒𝑛𝑡𝑠
Contd..

𝐶𝑂𝑀 = 0.28 (𝐹𝐶𝐼) + 2.73 (𝑜𝑝𝑒𝑟𝑎𝑡𝑖𝑛𝑔 𝑙𝑎𝑏𝑜𝑟 𝑐𝑜𝑠𝑡) + 1.23 (𝑐𝑜𝑠𝑡 𝑜𝑓 𝑢𝑡𝑖𝑙𝑖𝑡𝑖𝑒𝑠
+ 𝑐𝑜𝑠𝑡 𝑜𝑓 𝑟𝑎𝑤 𝑚𝑎𝑡𝑒𝑟𝑖𝑎𝑙)

𝑃𝑟𝑜𝑓𝑖𝑡 = 𝑅𝑒𝑣𝑒𝑛𝑢𝑒 – 𝐶𝑂𝑀

𝐻𝑒𝑎𝑡 𝑑𝑢𝑡𝑦 = 𝐻𝑒𝑎𝑡 𝑑𝑢𝑡𝑦

𝑎𝑙𝑙 𝑒𝑞𝑢𝑖𝑚𝑒𝑛𝑡𝑠

𝑂𝑟𝑔𝑎𝑛𝑖𝑐 𝑤𝑎𝑠𝑡𝑒
= 𝑐𝑜𝑚𝑝𝑜𝑠𝑖𝑡𝑖𝑜𝑛 𝑜𝑓 𝑚𝑒𝑡ℎ𝑎𝑛𝑜𝑙, 𝑔𝑙𝑦𝑐𝑒𝑟𝑜𝑙, 𝑢𝑛𝑐𝑜𝑛𝑣𝑒𝑟𝑡𝑒𝑑 𝑜𝑖𝑙 𝑎𝑛𝑑 𝐹𝐴𝑀𝐸
𝑖𝑛 𝑎𝑙𝑙 𝑒𝑥𝑖𝑡 𝑠𝑡𝑟𝑒𝑎𝑚𝑠
 Contd..
Simulation validation
Parameters Results of present Results reported by Results reported by Results reported by
simulation Sharma and Zhang et al. (2012) Garcia et al. (2010)
Rangaiah (2013)

Capacity 120 kt/annum 20 kt/annum 200 kt/annum 73.5 kt/annum

Thermo-physical property UNIF-DMD UNIQUAC UNIQUAC NRTL and UNIQUAC
method
Type of oil WCO WCO Pure oil Palm, Soyabean, WCO
FFA content (wt%) 6 5 0 5
Lumped/ Detailed Detailed Lumped Lumped Detailed
constituents
Esterification; 1;3 1; 3 0; 3 1; 2
transesterification
reactor(s)
Intermediate phase Yes Yes Yes No
separator in
transesterification
Washing scheme Methanol removal Methanol removal Methanol removal Washing followed by
followed by washing followed by washing followed by washing methanol removal
Purity of biodiesel (w/w 99.3 >99 >99 >98
%)
Purity of glycerol (w/w %) 96.5 >95 >99 19.71
Overall conversion (%) 99.6 >99 >99 Up to 94
Contd..
EMOO
(a) (b) (c)
12.0
+ 120 65

11.5

Heat Duty (MW)

WCO (kt/annum)
112

TRFFA ( C)
11.0

o
60

10.5 104

10.0 96 55
46 48 50 52 54 56 46 48 50 52 54 56 46 48 50 52 54 56
(d) (e) (f)
60 60 2.5

(Residence Time)RFFA (h)

TRTRANS2 (oC)
(o C )

55 55
2.0
TRTRANS1

50 50

45 45 1.5
46 48 50 52 54 56 46 48 50 52 54 56 46 48 50 52 54 56
(Residence Time)RTRANS1 (h)

(g) (h) (i)

2.5 10 9
9
(Feed Stage)FRAC-3

(Feed Stage)FRAC-4
8
8
7
7
2.0 6 6
5 5
4 4
3 3
1.5
2 2
46 48 50 52 54 56 46 48 50 52 54 56 46 48 50 52 54 56
Profit (million USD/annum) Profit (million USD/annum) Profit (million USD/annum)

Simultaneous maximization of profit and minimization of heat duty

Contd..
EMOO
(a) (b) (c)
0.21
120 65

Waste (kt/annum)

WCO (kt/annum)
0.20
112

TRFFA ( C)
o
60
0.19
104

0.18 96 55
48 50 52 54 56 48 50 52 54 56 48 50 52 54 56
(d) (e) (f)
60 60 2.5

(Residence Time)RFFA (h)

(o C )

TRTRANS3 (oC)
55 55
TRTRANS1

2.0
50 50

45 45 1.5
48 50 52 54 56 48 50 52 54 56 48 50 52 54 56
(Residence Time)RTRANS3 (h)

2.5 (g) (h) (i)

10 9
(Feed Stage)FRAC-3

(Feed Stage)FRAC-4
8
8
7
7
2.0 6
6
5 5
4 4
3 3
1.5
2 2
48 50 52 54 56 48 50 52 54 56 48 50 52 54 56
Profit (million USD/annum) Profit (million USD/annum) Profit (million USD/annum)

Simultaneous maximization of profit and minimization of organic waste

Contd..
Results of EMOO at ‘+’

Quantity Biodiesel process

FCI or CGR (million USD) 12.95

Methanol flow rate (kt/annum) 14.12

Biodiesel produced (kt/annum) 121.09

COM (million USD/annum) 73.5

Organic waste (kt/annum) 0.221

Heat Duty (MW) 11.75

Profit (million USD/annum) 55.44

Contd..
Biodiesel quality
Parameters Biodiesel EN14214
process
Methyl-O [wt%] 38.01 -
Methyl-P [wt%] 48.79 -
Methyl-M [wt%] 2.10 -
Methyl-S [wt%] 2.70 -
Methyl-Li [wt%] 7.69 -
Biodiesel (= sum of above)[wt%] 99.29 > 96.5

Monoglycerides [ppm] 286 < 8000

Diglycerides [ppm] 99 < 2000
Triglycerides [ppm] 112 < 2000
Glycerol [ppm] Trace < 2500
Methanol [ppm] 608 < 2000
Density (15°C) [kg/m3] 869.42 860-900
Contd..
Optimal DVs
Optimal DVs
Optimum TRFFA (oC) 65
Optimum TRTRANS1 (oC) 60
Optimum TRTRANS2 (oC) 55
Optimum TRTRANS3 (oC) 60
Optimum (Residence time)TRFFA (h) 1.7
Optimum (Residence time)TRTRANS1 2.1
(h)
Optimum (Residence time)TRTRANS2 2.1
(h)
Optimum (Residence time)TRTRANS3 2.1
(h)
Optimum (Feed Stage)FRAC-1 5
Optimum (Feed Stage)FRAC-2 7
Optimum (Feed Stage)FRAC-3 8
Optimum (Feed Stage)FRAC-4 7
Contd..

Example
EMOO of Biodiesel Process
 Summary

o (Multiobjective) optimization: need, methods and

application

o Application to an important biodiesel process

o Two bi-objective problems are solved using EMOO

program incorporating NSGA-II to investigate the
process performance.
Contd..

o Trade-offs between profit and heat duty, and profit and

organic waste are analyzed, which examine profitability
and environmental impact of the biodiesel process.

o Pareto-optimal fronts are obtained using EMOO.

o Optimal design and operating parameters have been

found.