Energy Demand Forecasting

Associate Professor Dr. Md. Hasanuzzaman

Higher Institution Centre of Excellence (HICoE)
UM Power Energy Dedicated Advanced Centre (UMPEDAC)
Room 16, Level 18, Wisma R&D, University of Malaya
59990 Kuala Lumpur, Malaysia
Tel:+603-22463405(Direct);+603-22463246(General Office)
Email: [email protected]; [email protected]

1
 Outline of Lecture
Understanding forecasting
Important of forecasting
Energy demand forecasting in Malaysia
Forecasting Accuracy
Methods of Forecasting
Forecasting Model Development
Forecasting Challenges

2
 Forecasting
 Using the past data / information to predict the future
 Taking historical data and projecting them into the future with some sort of
mathematical model (not an exact science: use of experience, judgement,
technical expertise, etc)
 Economic, technological and demand forecasts (i.e. demand and
availability of energy)
 What is the next number in the following sequences?
1, 4, 9, 16, 25, 36, 49, ....

3
 Why forecasting important?
Planning and operational decision conducted by utility companies (long range
planning)
Finance and accounting (budgets and cost controls)
Supply and demand fluctuating, change of weather conditions, prices increasing
during peak situations.
Short-term load forecasting can help to estimate load flows and to make decisions
that can prevent overloading.
Timely implementations of such decisions lead to the improvement of network
reliability and to the reduced occurrences of equipment failures and blackouts.
Evaluation of contract and various sophisticated financial products on energy pricing
offered by the market.
Decision on capital expenditures based on long-term forecasting of projects.
 Steps in Forecasting
Determine the purpose of the forecasting; items to be forecasted
Establish the time horizon
Select an appropriate forecasting model(s)
Gather and analyses the appropriate data
Prepare the forecast
Monitor the forecast

5
 Energy Demand Forecasting
 Energy demand forecasting is estimating future consumptions based on various data and
information available as per consumer behavior (i.e. daily forecast, monthly forecast, yearly or
multiyear forecast)
 Forecasting is an important issue of economic and safe operation planning in power distribution
systems.
 There are mostly three categories of energy demand forecasting:
short-term forecasting (an hour to a week) is mostly used for scheduling and analyses of the
distribution network.
Mid-term forecasting (a month to 5 years) is mostly applied for planning the power production
resources and tariffs,
Long-term forecasting (5–20 years) is mostly applied for resource management and
development investments.
 As energy demand changes by different time, climate variables, socioeconomic and demographics
parameters, accurately forecasting the demand is very important and difficult task
6
 Energy Demand Forecasting
 According to the International Energy Outlook
2016 (IEO2016), world net electricity generation
will be increased 69% by 2040, from 21.6 trillion
kilowatthours (kWh) in 2012 to 25.8 trillion kWh in
2020 and 36.5 trillion kWh in 2040.
 Coal share of total generation declines from 40%
in 2012 to 29% in 2040, even as world coal-fired
generation increases by 25% through 2040.
 At the same time, the shares of total generation
for both renewable energy sources and natural
gas expand: from 22% in 2012 to 29% in 2040 for
renewables and from 22% in 2012 to 28% in 2040
for natural gas.
Figure : World net electricity generation by fuel, 2012-40
(trillion kilowatthours)

7
Energy Demand Forecasting in Malaysia

Figure: Energy demand forecasting in Malaysia

(Peninsular Malaysia Electricity Supply Industry Outlook 2016) 8
Energy Demand Forecasting in Malaysia

9
Energy Demand Forecasting in Malaysia

10
Forecasting Accuracy

11
 Forecasting Accuracy
Forecasting needs to be done of unconstrained load
Forecasting involves both profile forecasting and total forecasting.
Profile forecasting is done more accurately while total forecasting is done less accurately
The forecasting accuracy is a combination of good data, good process and good model.
A good model alone or a good software alone can not give good accuracy
Accuracy improvement is a gradual process and involves significant human intervention.
More data does not necessary mean better forecast. It is important to select optimum
data size

12
 Methods of Forecasting
End Use Forecasting
Econometric Forecasting

13
 End Use Forecasting
 End use forecasting looks at individual devices, at end uses (e.g. cars)
How many cars?
How much fuel does a car use?
How will the number of car change in the future?
How will the amount of use per car change in the future?
Repeat for other end uses

 The relationships can be calculated via any of the methods used in time
series forecasting.
– Can be linear, polynomial, logarithmic…
 Relationships are determined simultaneously to find overall best fit.

14
 End Use Forecasting
Account for changes in efficiency levels (new cars tend to be more efficient
than older ones)
Allow for impact of competing fuels (natural gas vs. electricity for heating)
or for competing technologies (electric resistance heating vs. heat pump)
Incorporate and evaluate the impact of demand-side
management/conservation programs
Tremendously data intensive
Primarily limited to forecasting energy usage, unlike other forecasting
methods

15
 Econometric Forecasting
Econometric models attempt to quantify the relationship between the parameter of
interest (output variable) and a number of factors that affect the output variable.
 Example
– Output variable
– Explanatory variable
• Economic activity
• Weather
• Electricity price
• Natural gas price
• Fuel oil price
Each explanatory (independent) variable affects the output variable in different ways.
The relationships can be calculated via any of the methods used in time series
forecasting.
– Can be linear, polynomial, logarithmic…
Relationships are determined simultaneously to find overall best fit.
16
 Econometric Forecasting
Use historically established relationships between energy demand
and economic variables (e.g., GDP, population, household income)
Follow growth of driving variables (i.e., economic variables)
Estimation are made at more aggregated level or at sectoral level but
not at end-use level
They are unable to account technology specific features which are
key determinants of fuel consumption

17
Forecasting Error
Forecast is driven by a
number of assumptions
Usually impossible to
perfectly estimate the
relationship between all
possible factors and the
output
Uncertainty surrounding
long-term forecasts

18
Forecasting Model Development

20
 FORECASTING USING INDICATORS
Simple indicators commonly used for forecasting are:
Growth rates
Specific consumption
Energy intensity
Find the growth trend by fitting a time trend line is commonly used.
These methods is based on extrapolation or arbitrary assumption.
 TREND ANALYSIS
The trend analysis extrapolates the past growth trends and is normally
done by fitting some form of time trend to past behaviour.
The analysis:
Assumes that there will be little change in the growth pattern or in the
determinants of demand such as incomes, prices, consumer tastes, etc.
Finds the best trend line that fits the data. This is usually estimated by a least
square fit of past consumption data or by some similar statistical methodology.
The fitted trend is then used to forecast the future. Frequently, adjustments are
made to account for substantial changes in expected future demands due to
specific reasons.
 TREND ANALYSIS
 Linear Trend
o Fit the best straight line to the historical data and assume that the future
will follow that line
o Many methods exist for finding the best fitting line, the most common is
the least squares method.

Figure: Trend approach of time series energy demand model 23

 TREND ANALYSIS
 Polynomial Trend
o Fit the polynomial curve to the historical data and assume that the future will follow that line
o Can be done to any order of polynomial (square, cube, etc) but higher orders are usually
needlessly complex

 Logarithmic Trend
o Fit an exponential curve to the historical data and assume that the future will follow that line
These may not work well when there is a lot of variability in the historical data
If the time series curve does not perfectly fit the historical data, there is
model error.
There is normally model error when trying to forecast a complex system.

27
 TREND ANALYSIS
Depending on the availability of data, the analysis can be:

 Performed at the national level for a given energy source or they may be broken
down by region, by consuming sector or by both.

Used on its own or in combination with another method.

For example, if energy demand is estimated using per capita consumption (i.e.
unit consumption approach), the trend of population growth and per capita
consumption can be estimated using trend analysis.

The results can then be used in the unit consumption approach to get the final
results.
 TREND ANALYSIS
This is the most commonly used approach for forecasting and appears
to be more useful at relatively aggregate levels.
The simplicity of use is its main advantage.
It can be applied at aggregate and disaggregate levels and can be based
on whatever data is available.
Its disadvantages include:
future demand cannot be expected to depend on the past trend;
 it takes insufficient account of structural changes;
 it does not explain what determines demand as it does not explicitly include variables on price,
income, etc.
 TREND ANALYSIS
• Example: Using the global primary energy
consumption data available from the BP Statistical
Review of World Energy 2010, forecast the global
primary energy demand using the trend method.
• The original data and the fitted linear trend lines
are shown in Fig.
• The linear relationship is also shown in the figure.

• Using the forecast year as X in the above relation, Fig.: Trend analysis of global PEC
the forecast can be obtained.
Table : Forecast of global PEC using
• The fitted trend has achieved a good level of fit as trend line
is evident from the R2 information.
• The demand forecast for the period up to 2010
using this trend is given in Table.
 DIRECT SURVEYS
Direct surveys are generally used to generate primary information essentially for
the short term but surveys can also be used as a direct and reliable tool for
demand analysis and forecasting.
by analyzing the investment plans and programs, changes in the supply and
demand are captured.
But as surveys are costly undertakings, their use is generally limited to major
energy consumers such as medium to large size industrial plants, mines and
smelters, large transportation companies, utility companies, important
governmental users, etc.
The effects of new economic development programs, such as industrial
settlement, mining and hydrocarbon developments, etc. could also be assessed
using surveys but care must be taken to assess the realism of these specific
projected development programs.
While industrial surveys are more common, household energy surveys are less
common because of cost and time implications.
 DIRECT SURVEYS
The major problems with surveys as a tool for demand forecasting are as
follows:
This is a time consuming process and consequently involves high cost.
 It requires skilled staff to undertake the survey and analyze the results;
 As the survey depends on information provided by the respondents, the
quality of responses influences the results.
 It is quite possible that respondents are unwilling to divulge (hide)
information and can provide inaccurate information deliberately.
Many may not know the correct answers to the questions as well.
Relying on this information for forecasting could result in over/under
capacities.
 Forecasting Challenges

• Data / information related challenges

• Lack of good data is the biggest challenge in demand forecasting
• Process and Methodology Challenges
• Technology Challenges

42
Reference
 Hasanuzzaman, M., Rahim, N.A (2020) Energy for Sustainable Development, Demand, Supply, Conversion and
Management, 1st Edition, Elsevier, ISBN: 9780128146453, Pages 218, https://www.elsevier.com/books/energy-for-
sustainable-development/hasanuzzaman/978-0-12-814645-3.
 Bhattacharyya S. C., Energy Economics- Concept Issue, Markets and Governance, Springer London Dordrecht
Heidelberg, New York, 2011.
 Chong, C., W. Ni, L. Ma, P. Liu, and Z. Li, The Use of Energy in Malaysia: Tracing Energy Flows from Primary
Source to End Use. Energies, 2015. 8(4):2828.
 Malaysia Energy Statistics Handbook 2015, Energy Commission, Malaysia
 Dr. Ron Lembke, Energy: Supply and demand, business.unr.edu/faculty/ronlembke/352/ppt/02-4-Energy-2015.pptx
 Govinda R. Timilsina, Sectoral Models for Energy and Climate Policies, The World Bank, Washington, DC, Skopje,
Macedonia, March 1, 2011
 Atul Agrawal, Short Term Load Forecasting, Power Procurement Strategy and Power Exchanges, 28-30 July, 2014,
Indian Institute of Technology Kanpur (IITK) and Indian Energy Exchange (IEX) are delighted to announce,
 Douglas J. Gotham, Energy Forecasting Methods, State Utility Forecasting Group, Energy Center Purdue
University, November 15, 2007
 Peninsular Malaysia Electricity Supply Industry Outlook 2016, Energy Commission, Malaysia

43