Portfolio selection for energy projects under the Clean

Development Mechanism (CDM)

Olena Pechak, PhD candidateGeorge Mavrotas, Asst. Professor

School of Chemical Engineering

National Technical University of Athens, Greece

1ο ΠΑΝΕΛΛΗΝΙΟ ΦΟΙΤΗΤΙΚΟ ΣΥΝΕΔΡΙΟ ΕΛΛΗΝΙΚΗΣ ΕΤΑΙΡΕΙΑΣ ΕΠΙΧΕΙΡΗΣΙΑΚΩΝ ΕΡΕΥΝΩΝ (Ε.Ε.Ε.Ε.)

E.E.E.E., 25 – 27 Νοεμβρίου 2010 2

Structure of the presentation

Introduction CDM projects Project selection problem Description of the method

Two step method Incorporate uncertainty

Monte Carlo simulation in GAMS Results and conclusions

E.E.E.E., 25 – 27 Νοεμβρίου 2010 3

IntroductionClimate change is a complicated problem created by humanity. In

order to face it, UNFCCC was adopted. It coordinates activities for mitigation and adaptation to climate change.

Kyoto Protocol to the UNFCCC provides three types of flexible mechanisms to reduce GHG emissions:

joint implementation (JI), clean development mechanism (CDM) and international emission trading (IET).

RES activities within CDM: 65% of all projects in CDM pipeline 46% of emission cuts

858 wind energy projects (17.4% of total in Jan. 2010) 284 of which (with installed capacity - 12.55 GW) are

already registered.

E.E.E.E., 25 – 27 Νοεμβρίου 2010 4

Introduction: Main features of CDM

Cooperation between developing and developed countries

The projects activity provides GHG emission reductions compared with BAU scenario,

During the operation of the project achieved reductions are translated into Certified Emission Reductions (CERs), backed by the 1 t CO2-eq

CERs may be sold in carbon market. The price for them is variable

Project duration is variable and is chosen before the registration of the project

E.E.E.E., 25 – 27 Νοεμβρίου 2010 5

Introduction: Wind energyInfluencing factors on

popularity of wind energy:

growing energy demand,

increased concerns for environmental and climate issues,

improvements of technology itself.

3 key regions:Europe,North America, andAsia (with China and India as main players)

E.E.E.E., 25 – 27 Νοεμβρίου 2010 6

Project selection problem

Selection of project portfolio with constraints like: Policy Budget Geographical distribution Technical constraints and MW of installed capacity Logical constraints

In addition we have: Set of hypothetic projects Variable CER prices

Solving software: GAMS CPLEX 12.2

E.E.E.E., 25 – 27 Νοεμβρίου 2010 7

Project selection problemInput 100 CDM projects across 4 regions (China 50, India

30, Latin America 10, Mediterranean 10) Budget constraint is 2.8 billion $, the sum of

candidates - 4.2 billion $ regional constraints (at least 3 projects from Latin

America and Mediterranean, India must have no less than half of China)

MW constraints - not more than 75% of MW in China logical constraints (mutually exclusive projects) technology constraints (sum of off shore MW across

all countries less than 2 GW)

E.E.E.E., 25 – 27 Νοεμβρίου 2010 8

Project selection problemIn calculations we: maximize total NPV (with discount rate =8%) consider only the CER price as uncertainty factor perform Monte Carlo simulation and optimization

for normal distribution around 20$ (μ=20, σ=3.3)

for uniform distribution in [10, 30]

The process of solving the problem is divided into 2 big steps.

E.E.E.E., 25 – 27 Νοεμβρίου 2010 9

Description of the method

Step I An Integer Programming (IP) model is developed with

binary decision variables that express the existence of each project in the selected portfolio. objective function - cumulative NPV

constraints: policy, logical and budget Due to the uncertainty related to the future price of

Certified Emission Reductions (CERs) a Monte Carlo simulation-optimization process is designed. Projects from resulting portfolios form subsets: Green set – projects are present in all portfolios Red set – none of projects is selected in any

portfolio Grey set – projects, present in some portfolios

E.E.E.E., 25 – 27 Νοεμβρίου 2010 10

Description of the method

Step II

Only projects from “grey” set are under consideration A new IP is formulated only with the “grey” projects

and uses the frequency obtained from the Monte Carlo simulation as their objective function coefficients.

The constraints are the same as before appropriately modified to take into account the sure adaptation of the “green” projects. The final output is the set of projects (portfolio) with the best performance under the given uncertainty conditions.

Graphical representation of Monte Carlo simulation and optimization

A B

μ

CER price(i)

CER price(i)

Solution of IP

i =1…1000

NPV

0

5

10

15

20

25

30

35

-1.500.000

-1.122.000

-744.000

-366.000

12.000

390.000

768.000

1.146.000

1.524.000

1.902.000

2.280.000

2.658.000

3.036.000

3.414.000

3.792.000

4.170.000

4.548.000

Obj function (z)

unif

orm

norm

al

E.E.E.E., 25 – 27 Νοεμβρίου 2010

E.E.E.E., 25 – 27 Νοεμβρίου 2010 12

Results

The normal and uniform distributions of projects’ NPV during the Step I

E.E.E.E., 25 – 27 Νοεμβρίου 2010 13

ResultsResults of the Step I in both cases are similar. The differences

are observed only in frequencies of the projects from the “Grey” set.

Totals: Green set – 58 Red set – 23 Grey set – 19

GREEN set

1,3,4,6,8,9,10,11,13,14,15,16,17,18,21,23,24,25,26,28,30,37,40,42,44,46,47,49,50,52,53,55,56,57,58,61,63,65,66,71,72,74,75,77,78,79,81,82,83,86,87,88,90,91,92,93,94,97

RED set

2,5,12,19,20,22,27,31,33,35,36,41,45,48,54,59,68,73,89,95,96,99,100

GREY set

7,29,32,34,38,39,43,51,60,62,64,67,69,70,76,80,84, 85,98

E.E.E.E., 25 – 27 Νοεμβρίου 2010 14

ResultsLet’ s see the Grey sets more carefully.

# 7 29 32 34 38 39 43 51 60

Normal 36 969 317 988 796 552 294 192 796Uniform 28 787 566 868 626 233 416 463 405

# 62 64 67 69 70 76 80 84 85 98

N 149 677 770 404 988 860 988 988 988 57

U 344 557 611 488 868 516 868 868 868 62

E.E.E.E., 25 – 27 Νοεμβρίου 2010 15

Results: final selectionResults of the Step II.Case of Normal distribution of CER prices:

Case of Uniform distribution of CER prices:

Projects Total MW Budget

1,3,4,6,8,9,10,11,13,14,15,16,17,18,21,23, 24,25,26,28,29,30,32,34,37,39,40,42,44, 46, 47,49,50,51,52,53,55,56,57,58,60,61,63,64, 65,66,67,69,70,71,72,74,75,76,77,78,79,80, 81,82,83,84,85,86,87,88,90,91,92,93,94,97,98

73 2612 2.789Billion $

Projects Total MW Budget

1,3,4,6,8,9,10,11,13,14,15,16,17,18,21,23, 24,25,26,28,29,30,32,34,37,40,42,43,44, 46, 47,49,50,51,52,53,55,56,57,58,60,61,63,64, 65,66,67,69,70,71,72,74,75,76,77,78,79,80, 81,82,83,84,85,86,87,88,90,91,92,93,94,97,98

73 2631 2.799 Billion $

E.E.E.E., 25 – 27 Νοεμβρίου 2010 16

Conclusions about the method

Integer Programming can be effectively used for a portfolio selection problem

The objective function can be selected in order to reflect the decision making preferences

The combination of Monte Carlo and optimization is used successfully for dealing with uncertainty

The two step approach is a useful decision aid tool that (a) classifies the projects and (b) uses the information from the first step to drive the second optimization and result in the final choice

E.E.E.E., 25 – 27 Νοεμβρίου 2010 17

Conclusions about the problem The resulting sets of projects from the Monte Carlo

simulation are the same for normal and uniform distribution.

The difference between uniform and normal distribution is observed in frequencies of “grey” projects. It also influences the ranking of the grey projects between each other.

19 ambiguous projects (grey set) from all kinds

Normal and uniform distribution give almost the same final choice

In final selection there are: China – 33 projects (#39 and #43 are very similar), Latin America – 8 projects, Mediterranean – 8 projects, India – 24 projects.

E.E.E.E., 25 – 27 Νοεμβρίου 2010 18

Future research

Incorporate multiple criteria MCDA and then IP

Multiobjective IP

Group decision making IP for optimization

Multiple portfolios, one for each DM

Same approach (green, red, grey set) Step1: Green set the principle of unanimity

Step2: Grey set the principle of majority

E.E.E.E., 25 – 27 Νοεμβρίου 2010 19

Thank you!