You are on page 1of 38

Predicting the Electricity Demand

of Turkey Using the Past Data


Assist. Prof. M. Erdem Gnay
Energy Systems Engineering
Istanbul Bilgi University
erdem.gunay@bilgi.edu.tr

Energy and Society

ENERGY AND SOCIETY

The

more a country is developed the


more energy it consumes.
Population and wealth of a country
directly affect the energy consumption

List of countries by population

Ref: http://en.wikipedia.org/wiki/List_of_countries_by_population

Map of Countries:
Gross Domestic Product

GDP is the addition of the consumer spending, investment spending, government


spending and the value of the exports and subtract the value of the imports.

This measure is often used to find out the health of a country in an economic way.

A country with a high value of GDP can be called a large economy.

Ref: http://simple.wikipedia.org/wiki/Gross_domestic_product

Map of Countries:
CO2 emissions

The more energy is consumed the more pollutants are


released.

Ref: http://en.wikipedia.org/wiki/List_of_countries_by_carbon_dioxide_emissions

ENERGY AND SOCIETY

The developed countries consume the most energy


and produce the most pollution, primarily due to
the increase in the amount of energy per person.

Ref: Introduction to Renewable Energy, Abbas, Ghassemi, CRC Press

Turkey is getting more populated


>20 Million
1950

Turkey is getting more populated


>30 Million
1960

Turkey is getting more populated


Population (Million People)

80

70

>50 Million

60
50
40

1985

30
20
10

0
1935

1955

1975
Year

1995

2015

Turkey is getting more populated


~80 Million

Population (Million People)

80

70
60

2014

50
40
30
20
10

0
1935

1955

1975
Year

1995

2015

Gross Domestic Product Per Capita of


Turkey is Also in an Increasing Trend

GDP per capita (Int $)

20,000

15,000

10,000

5,000

0
1975

1985

1995
Year

2005

GDP per
capita is an
indicator
of the
average
living
standards

2015

Energy Demand of Turkey

~250 TWh

Electricity Demand (TWh)

300

~20 TWh

250
200

150
100

50

1975

1985

1995
Year

2005

2015

The energy demand of Turkey increases year by year as


the population rises and as we become more and more
dependent on power hungry machines and devices.

Energy Supply of Turkey in 2013


Natural Gas
4% 2%
Coal
25%

44%
Hydroelectricity

25%
Renewable
Energy
Liquid Fuels

Reference: Ministry of Energy and Natural Sources Web Page (www.enerji.gov.tr)

Foreign Dependence of Energy of Turkey


% Dependence
on Foreign Countries

Reference: Ministry of Energy and Natural Sources Strategic Plan 2010

Energy Demand of Turkey


Problem:
The energy demand of Turkey is expected to increase
in an accelerating trend in the future.

Energy sources satisfying this


demand will become totally
insufficient.

The foreign dependence of


energy will even become more
serious in the future

Energy Demand of Turkey


Solution:
Investments on alternative energy sources must be encouraged.
The local resources must be used more effectively
The types of energy sources must also be increased.

We need to know the exact energy


demand in the future years for these
plans to be effective

We should predict the future energy


demand from the historical data
with a good accuracy.

Predicting the Energy Demand of Turkey


What happens if we cannot predict the future
electricity demand accurately?

Underestimation of the
demand would lead to
blackouts, economic crisis etc.

Overestimation would lead to


unnecessary idle capacity that
means wasted financial resources.

Predicting Energy Demand for the Future


Which Factors May Affect the Demand?

We know that
Population and
wealth of a
country directly
affect the
energy
consumption.

Predicting Energy Demand for the Future


Which Factors May Affect the Demand?

Inflation?
It may have an
effect on the
energy demand.

Predicting Energy Demand for the Future


Which Factors May Affect the Demand?

Summer
average
temperature?
The hotter it is
in the summer
the more we
use air
conditioners.

Constructing
the
Database

Turkish
statistical
institute

Year Population
1975
1976
1977

2008
2009
2010
2011
2012
2013
2014
2015

2020

40347719
41225567
42103414

71517100
72561312
73722988
74724269
75627384
76667864
?
?
?
?
?

The Data
OECD: Organisation for
Economic Co-operation
and Development

Turkish State
Meteorological Service

GDP per Inflation Avg T Sum


capita ($)
(%)
(oC)
2020
2306
2481

15021
14550
16003
17781
18315
18834
?
?
?
?
?

16.26
19.33
46.77

10.06
6.53
6.40
10.45
6.16
7.40
?
?
?
?
?

23.13
21.64
23.03

24.64
23.78
25.22
23.87
24.69
23.85
?
?
?
?
?

Turkish Electricity
Transmission Company

Demand
(TWh)
15.7
18.6
21.1

198.1
194.1
210.4
230.3
242.4
245.5
Unknown
Unknown
Unknown
Unknown
Unknown

Used for
training
(constructing
the model)
Used for validation
(checking whether
the model works
or not)

Used for Testing


(predicting
energy demand
for the future)

Step 1: Constructing the Model:


Linear Multiple Regression to Model the
Demand between the Years 1975-2008
Demand a0 a1 pop a2 gdp a3 inf a4 Ts

Demand (TWh)
Pop: Population
Gdp: gdp per capita (Int $)
Inf: Inflation (%)
Ts: Average summer temperature (oC)

Advantage of Multiple Regression:


It is not a black box modeling i.e artificial neural
networks.
After determining the coefficients we can use the
equation for predicting the demand of any year.

Modeling the Demand between for the Years 1975-2008

Variables
constant
Population
GDP
inflation
Ts

Standard
Error of the
Coefficient Coefficient t value
-198
55.9
-3.54
1.60E-06
4.22E-07
3.80
0.010
1.35E-03
7.30
-0.19
0.0552
-3.42
5.58
2.51
2.22

p value
0.0014
0.00069
0.00000
0.0019
0.034

The two-tailed P values are smaller


than 0.05
By conventional criteria, all the
variables are considered to be
statistically significant.

Modeling the Demand between for the Years 1975-2008

Demand 198 1.60 10 pop


0.010 gdp 0.19 inf 5.58 Ts

Demand (TWh)
Pop: Population
Gdp: gdp per capita (Int $)
Inf: Inflation (%)
Ts: Average summer temperature (oC)

Step 2:Validating the Model:


Linear Multiple Regression to Model the
Demand between the Years 2009-2013

Demand 198 1.60 106 pop


0.010 gdp 0.19 inf 5.58 Ts
Year Population
2009
2010
2011
2012
2013

72561312
73722988
74724269
75627384
76667864

GDP per Inflation Avg T Sum


capita ($)
(%)
(oC)
14550
6.53
23.78
16003
6.40
25.22
17781
10.45
23.87
18315
6.16
24.69
18834
7.40
23.85

Demand
(TWh)
194.1
210.4
230.3
242.4
245.5

Step 2:Validating the Model:


Linear Multiple Regression to Model the
Demand between the Years 2009-2013

We normally know the pop., gdp, inf. and Ts,


between the years 2009-2013.
The problem is, we do not know the values
of these in the future.
If we can correctly estimate them between
the years 2009-2013, then we can also
predict them for the future years.
We need another tool to predict the
corresponding values.

Autoregressive Regression
The autoregressive regression model specifies that an
output variable depends on its own previous values.

Population(t ) f Popl.(t 1), Popl.(t 2), Popl.(t 3)...


For example the population in the year 2014 depends on
the population 2013, 2012, 2011 etc.

GDP (t ) f GDP (t 1), GDP (t 2), GDP(t 3)...


The same is true for GDP and other variables.

Autoregressive Regression to Predict the


Population Between the Years 2009-2013
Population(t ) 408779 1.01 Population(t 1)
Real

Estimated

Population (Millions)

80

Predicted

R2predicted=0.995

60
R2estimated=1.000

40
20
0
1935

1955

1975
Year

1995

2015

Autoregressive Regression for to Predict the GDP


per Capita Between the Years 2009-2013

GDP (t ) 7.97 1.06 GDP(t 1)


Real

Estimated

Predicted

GDP per capita (Int $)

20,000
R2predicted=0.823

15,000

10,000

R2estimated=0.988

5,000
0
1975

1985

1995
Year

2005

2015

Autoregressive Regression to Predict the


Inflation % Between the Years 2009-2013
Inf (t ) 1.00 1.40 Inf (t 1) 0.0062 Inf (t 1)

Real

Estimated

Predicted

140

120
Inflation %

100
80

R2estimated=0.722

60
40

R2predicted=0.499

20
0
-201955

1975

1995
Year

2015

Autoregressive Regression to Predict the Average


Summer Temperature Between the Years 2009-2013

Ts(t ) 1.72 0.45 Ts(t 1) 0.22 Ts(t 2) 0.09 Ts(t 3)


0.11 Ts(t 4) 0.41 Ts(t 5) 0.05 Ts(t 6)
Average Summer Temperature (oC)

Real

Estimated

Predicted

25.5

R2predicted=0.215

25.0
24.5
24.0

R2estimated=0.597

23.5
23.0

22.5
22.0

21.5
1975

1985

1995
Year

2005

2015

Predicting the Electricity Demand


for the years Between 2009-2013
Year
2009
2010
2011
2012
2013

Population GDP per capita ($)


72561384
73615026
74678109
75750718
76832939

15645
16607
17627
18710
19859

Inflation
(%)
12.48
7.89
7.72
12.98
7.40

Avg T Sum
(oC)
24.29
24.21
24.32
24.61
24.72

Demand 198 1.60 10 pop


0.010 gdp 0.19 inf 5.58 Ts

Predicting the Electricity Demand


for the years Between 2009-2013
Real

Estimated

Predicted

Electricity Demand (TWh)

300
250

R2predicted=0.809

200
150
100

R2estimated=0.983

50

1975

1985

1995
Year

2005

2015

Predicting the Electricity Demand


for the years Between 2009-2013
Real Demand
EPDK High Demand Scenario
EPDK Low Demand Scenario

Real Demand

Predicted by the current model

290

270

R2=0
Average
% Error = 13.5%

R2=0
Average
% Error = 10.2%

250

230
210
190
2008

2010

2012
Year

Republic of Turkey Energy Market


Regulatory Authority Prediction
Overestimation

2014

Electricity Demand (TWh)

Electricity Demand (TWh)

290
270

R2=0.81
Average
% Error = 3.1%

250
230
210
190
2008

2010

2012
Year

Current Model

2014

Step 3: Using the Data of the Past to Predict


the Demand between the Years 2014-2020

The data between 1975 to 2013 were used


to construct a multiple regression model.
6

Demand 181 1.61 10 pop


0.010 gdp 0.19 inf 4.80 Ts

Population, gdp per capita, inflation and


average summer temperature were
predicted for the years 2014-2020.
These values were used to calculate the
energy demand.

Step 3: Using the Data of the Past to Predict


the Demand between the Years 2014-2020
Real

Estimated

Predicted

~320 TWh

Electricity Demand (TWh)

320

240
R2estimated=0.991
160

80

1975

1984

1993
2002
Year

2011

2020

You might also like