Demand models and forcasting Part Two.pdf

Forecasting

1. Engr. Dexrey John P. Quizan

2. Engr. Dexrey John P. Quizan
University of Perpetual Help System Dalta - Molino

NOTICE

The materials in this folder have been reproduced and communicated to you by or on behalf of the
Professor and/or University of Perpetual Help System Dalta- Molino Campus pursuant to PART IV: The
Law on Copyright of Republic Act (RA) 8293 or the “Intellectual Property Code of the Philippines”.

The Professor and/or University does not authorize you to reproduce or communicate this material.
The Material may contain works that are subject to copyright protection under RA 8293. Any
reproduction and/or communication of the material by you may be subject to copyright infringement
and the copyright owners have the right to take legal action against such infringement. Do not
remove this notice.

3. 3
FORECASTING MODELS
❑Growth rates
❑Econometric models
❑Trend models
❑Time series models
❑End-use models
❑Spatial load forecasting

4. 4
Outline
❑End-use models
�Residential end-use
�Commercial end-use
�Industrial end-use
�Transport end-use
�Econometric vs End-use models
❑Spatial forecasting models

5. 5
End-Use Models

6. 6
End-Use Models
❑also known as “end-use projection models” or
“engineering models”
❑a bottom-up approach to energy modeling
❑focus on the various uses of energy in the residential,
commercial, and industrial sectors
❑follow the intuitive notion that energy is used only as an
intermediate means of obtaining a desired energy service
(heat, light, power)

7. 7
Billion of kWh
Year
Composition of Residential Load Forecast
Residential End-Use Load Forecasting
Residential Appliance Saturation Method

8. 8
Year
Billion of kWh
Composition of Commercial Load Forecast
Commercial Building Saturation Method
Commercial End-Use Load Forecasting

9. 9
Billion of kWh
Composition of Industrial Load Forecast
Industrial End-Use Load Forecasting
Sectoral Energy Intensity Model

10. 10
End-Use Models –Basic Structure
Energy demand
= level of activity(energy service)
×energy intensity(energy use per unit of energy service)
Q
i= quantity of the energy service
I
i= intensity of energy use for
energy service i

11. 11
End-Use Model –Basic Structure
Q
i quantity of energy service i
N
i number of customers eligible for end use i(e.g., number of
households, commercial premises or industrial customers, total
amount of commercial floor space
P
i penetration of end use service i= share of eligible customers that
use a given energy service = total units/total customers → can be
> 100% for multiple units per customer
M
i magnitude or extent of use of energy use

12. 12
The level of energy service
Q
idepends on:
❑Economic activity of the customer class considered
❑Patterns of energy usage
❑Information on penetration levels of energy services
within the customer class
End-Use Model –Basic Structure

13. 13
End-Use Models –Data Requirements
●Equations used in the end-use forecasting approach
require a breakdown by sectors, activities, and end-uses
●Data sources:
○ Equipment saturation–aggregate indicators like appliance
sales; if not available, may opt to use existing info from other
countries with similar socio-economic development
characteristics
○Ownership and energy intensity–Questionnaire-based
surveys, billing data analysis, energy audits and measurements

14. 14
End-Use Models –Example 1
In a community of 100 households, 80% own a TV. The
average TV consumes 200 W of electricity and is turned on for
an average of 2 hours per day. What is the annual energy use
for televisions in the said community?

15. 15
End-Use Models –Example 1
Solution:
N
i= 100 households
P
i= 80% = 80 TV sets/100 households
M
i= 2 hrs/day × 365 days/year = 730 hrs/year
I
i= 200 W/TV = 0.2 kW/TV

16. 16
End-Use Models –Example 2
A store contains 1,000 m
2
of floor space, 90% of which is lit
by fluorescent lighting. Each 15 m
2
of floor area contains four
40 W fluorescent bulbs, each of which provides a lighting
output of 67 lumens/W. The lights are on for an average of 8
hours per day. What is the annual energy use for fluorescent
lighting for the store?

17. 17
End-Use Models –Example 2
Solution:
I
i= 1 watt/67 lumens = 0.0149 W/lumen = 0.000149 kW/lumen
N
i= 1,000 m
2
of floorspace
P
i= 90% = 90 m
2
lighted floorspace/100 sqm total floorspace

18. 18
Residential End-Use Load Forecasting
Demographics
Households
Housing
Appliance Saturation
House Simulation
Residential Forecast
Appliance kWh Usage
Residential Appliance Saturation Method

19. 19
Steps
1. Forecast number of households
2. Determine current saturation level of appliances
3. Forecast new electricity-consuming devices
4. Forecast future penetration of appliances
5. Determine electricity usage of appliances
6. Forecast future efficiency improvements of appliances
7. Compute resulting electricity demand (kWh)
8. Validate forecast
Residential End-Use Load Forecasting
Residential Appliance Saturation Method

20. 20
Refrigerator (in Millions) Example [U.S. Data]
Residential Appliance Saturation Method
Residential End-Use Load Forecasting

21. 21
Energy Forecast Energy Consumption
Refrigerator Example [U.S. Data]
Residential Appliance Saturation Method
Residential End-Use Load Forecasting

22. Families Income and
Expenditures Survey (FIES)
22

23. Households Energy
Consumption Survey
23

24. 24
Residential End-Use Model: Philippines Example
Links:
https://www.adb.org/sites/default/files/
publication/389806/pathways-low-
carbon-devt-philippines.pdf
http://moef.gov.in/wp-
content/uploads/2018/04/Residentialpo
werconsumption.pdf
Uses the Energy Forecasting
Framework and Emissions Consensus
Tool (EFFECT) model
Residential model and vehicle
ownership models based on a model
developed for India

25. 25
Econometric
Data
FIES

26. 26
Residential End-Use Model: Philippines Example
% electrified households
Monthly per capita expenditure
Sample regression model for electrification and appliance penetration

27. 27
Residential End-Use Model: Philippines Example
Sample survival curves for different average life span

28. 28
Residential End-Use Model: Philippines Example

29. 29
Residential End-Use Model
(Australia Example)
Prepared by the Department of Environment, Water, Heritage and Arts, 2008

30. 30
Modeling
overview

31. 31
Appliance Modeling Overview
Retirement function - Stock Model Stock Remaining - Stock Model
Graphical Depiction of the Stock model
Example stock average life of 10 years with a standard deviation of 2 years
Notes:
1.New appliance stock in 1993
persists until about 2012 (gray).
2.Appliance stock in 2006 is
composed of new appliances
purchased in 2006 and
previously installed appliances.
3.The characteristics of new
appliances may vary year to
year.

32. 32
Key Results
Trends in Total Energy Consumption by End-use

33. 33
Key Results
Trends in Electrical Appliance Energy by Type

34. 34
Demographics
Economics
Floor Space
End-Use Saturation
End-Use Energy Usage
Commercial Forecast
Commercial End-Use Load Forecasting
Commercial Building Saturation Method

35. 35
Steps
1. Classify building types
2. Classify end uses
3. Determine building floor area
4. Determine electric end-use penetration rate and annual
electricity usage per floor area
5. Forecast future building floor area additions
6. Forecast future end-use penetration of rate and annual
electricity usage per floor area
7. Compute resulting total energy consumption
8. Validate forecast
Commercial End-Use Load Forecasting
Commercial Building Saturation Method

36. 36
Commercial Floor Estimates (Millions Square Feet) [U.S. DATA]
Commercial Building Saturation Method
Commercial End-Use Load Forecasting

37. 37
Typical Commercial Annual kWh/ft2 End-Use Data [U.S. Data]
Commercial Building Saturation Method
Commercial End-Use Load Forecasting

38. Commercial End-Use Modeling
(California Example)
38

39. Key Results
39

40. Key Results
40

41. Key Results
41

42. 42
Economics
Electrification Trends
INDUSTRYMODELS
Industrial kWh
Industrial Forecast
Industrial End-Use Load Forecasting
Sectoral Energy Intensity Model

43. Industry Sectors
43
Industrial End-Use Load Forecasting
Sectoral Energy Intensity Model

44. 44
Total Industrial Production [U.S. Data]
Year
Growth Rate (%/Year)
Industrial End-Use Load Forecasting
Sectoral Energy Intensity Model

45. 45
Energy Intensity (kWh/Industrial production)
Mining & Manufacturing [U.S. Data]
Industrial End-Use Load Forecasting
Sectoral Energy Intensity Model

46. Transport Sector Disaggregation (1/2)
46
Transportation End-Use Load Forecasting

47. 47
Transport Sector Disaggregation (2/2)
Transportation End-Use Load Forecasting

48. Vehicle Ownership Models
○Car ownership follows an S-shaped curve when plotted
against income
○Functional forms:
48
Transportation End-Use Load Forecasting
C car ownership (e.g. vehicles per 1000 people)
GDP per capita income
S saturation level (vehicles per 1,000 people)
a, b define the shape or curvature of function

49. 49
Transportation End-Use Load Forecasting

50. Determinants of passenger transport energy demand
50
Transportation End-Use Load Forecasting

51. Determinants of freight transport energy demand
51
Transportation End-Use Load Forecasting

52. 52
https://www.adb.org/sites/default/files/publication/389806/pathways-low-carbon-devt-philippines.pdf
Transport End-Use Model: Philippines Example

53. 53
Econometric Models vs.
End-Use Models
●Are end-use models more accurate than econometric
models?
●Which model should be used: end-use models or
econometric models?

54. 54
Econometric Models vs.
End-Use Models
●Are end-use models more accurate than econometric
models?
○There is no inherent reason why one approach would be more
accurate than the other
○To the extent that econometric models are successful in
uncovering the underlying consumer behavior that relate key
influencing factors such as levels of economic activity, personal
income, demographic characteristics, energy prices, etc. to the
amount of energy consumed, these models should do a
commendable job of forecasting energy consumption
■A major question though is whether that underlying structure (e.g.,
income elasticity and price elasticity) remains the same or is, in fact,
changing in a way that the econometric model is unable to uncover

55. 55
Econometric Models vs.
End-Use Models
●Are end-use models more accurate than econometric
models?
○End-use models contain considerable structural detail and thus
can deal with structural changes. The problem in forecasting with
these models is that change in structure must be assumed or
itself forecasted in some way
○Ultimately, econometric and end-use models are driven by the
same projections of the influencing variables, such as rates of
economic growth and energy prices. Some studies show that
forecasts of these driving variables as important, if not more
important, than the energy forecasting models themselves.

56. 56
Econometric Models vs. End-Use
Models
●Which model should be used: end-use models or
econometric models?
○Clarification: End-use and econometric methods are
types of analysis rather than necessarily types of
forecasting models.
○The real issue: Use (a)
aggregateeconometric models
or (b) disaggregatestructurally detailed end-use
models?

57. 57
Econometric Models vs. End-Use
Models
●Which model should be used: end-use models or
econometric models?
○The choice among aggregate and disaggregate methods
or models should be made in the context of a definite
objective. For example:
■If the objective is the capability to turn around a forecast quickly and
have a model that is easily and transparently described in terms of
key sensitivities (e.g., elasticities), then an aggregate econometric
model may be appropriate
■If the need is to develop a detailed demand-side plan in which
various DSM strategies are evaluated and incorporated into the
forecast, use end-use models.

58. Engr. Dexrey John P. Quizan