Forecasting load demand is a difficult procedure and combines art with science.
While forecasting tools are available to aid in the process, the key contributions
of forecasters are their knowledge of electricity consumers and an understanding
of the way they use electricity and other competing energy forms. Overall demand
forecast framework and steps involved are shown in Figure 2.
Demand Forecasting Steps
STEP 1: Demand Forecasting Objective
The objective and scope of the forecasting exercise would be specific and different
for each of the stakeholders in power business and is prepared to meet their
requirements. The forecast prepared by the distribution entity could be drawn
from the individual consumer segment or end use forecasts. It gains significant
importance considering its relevance for the least cost integrated planning which
includes DSM options along with the utility supply options as well as to asses
the impact of environmental and social costs.
The demand forecasting exercise assumes greater significance in the context
of introduction of energy efficient technologies, awareness of environmental
social impact of energy, power exchange, competition and open access in the
Indian electricity market.
The basic “Demand Forecasting” objective from the perspective of
DSM is that the benefit of the electricity services should be measured in terms
of energy services provided, and not in terms of energy sold. Thus to assess
the DSM potential, scenario based forecasting approach is most suitable. Where
the growth in energy services is forecasted with and without energy efficient
STEP 2: Demand Forecasting Methods
Different methods are developed for forecasting the demand in the last two
decades. In the past, straight line extrapolations of the historical trends
purpose well. However, with the beginning of rapidly rising energy prices,
inflation, appearance of alternative fuels and DSM options, changes in the
energy consumption pattern etc, it is become important to use the modelling
techniques, to capture the effect of factors such as energy efficient technologies,
per capita income, population and other variables.
Some of the most commonly used methods for demand forecasting includes econometric
regression analysis, appliance saturation methods, end use energy methods,
time series / trend analysis etc. The usefulness of each method depends on
customer segmentation, and degree of detail required. An appropriate method
could be chosen based on the nature of the data available and the desired nature
level of detail of the forecasts. An approach often used by the forecaster
is to employ more than one method and then to compare the forecasts to arrive
a more accurate forecast.
The forecaster may use a combination of forecasting techniques that give him
aggregate annual forecasts and those that predict hour-by hour demand for electricity
in individual sectors. Central Electricity Authority (CEA) has been using partial
end-use method specific to different consumer categories to project demand
in different states. The time horizons in CEA’s case have been detail forecast
for 5 years which has been extrapolated for the next 5 years. Forecasts done
by power survey office of CEA are published as “Electric Power Survey”(EPS).
The methods commonly referred by utilities for demand forecasting are discussed
in detail in following sections. Most of these could be used for both long-term
as well as short-term forecasting.
Method 1 : Econometric Regression Analysis
Econometric regression analysis uses historical annual energy and economic
data to determine customer elasticities. Elasticity is measure of how a customer
will change a purchasing pattern in response to a change in price, convenience,
reliability and other factors. Based on customer elasticities and assuming
that that these elasticities do not change through time, demand forecast is
Econometric analysis combines economic theory with statistical methods to produce
a system of equations for forecasting energy demand. In this method of forecasting
the functional relationships are established between the dependent variable,
in this case the electricity demand (energy as well as peak) and explanatory
variables like GDP, energy price, energy efficient technologies etc. The relationship
is of the form,
ED=f(Y, Pi, Pj, Pop, T)
ED = Energy demand,
Y = Output or income,
Pi = Power Price,
Pj = Price of related fuels,
Pop = Population and
T = Technology.
The forecast of the explanatory variables is used in the functional relationship
to obtain the forecast of the dependent variable i.e. the energy demand. The
advantage of this method is that socio economic factors like population, income,
tariffs, technology diffusion etc. taken in to account while forecasting the
values of the dependent variable. The disadvantage is that econometric methods
require consistent data for a reasonably long period of time.
Method 2 : Appliance Saturation Method / End Use Energy Method
The appliance saturation method is an “engineering” type methodology.
Load research surveys are made to determine the number of customers with a
certain appliance (for example, a refrigerator) and the typical annual energy
used by the appliance. Then, on the basis of a forecast of the number of appliances
expected in the future, together with the forecast of how the annual energy
usage per appliance will change, the energy demand forecast is made.
The end use energy method is similar to the appliance saturation method, except
that instead of using an appliance as forecasting basis, the basis is the end
use process. The floor space and kilowatt-hour energy consumption of the principle
electric devices per square foot (space heating and cooling, lighting and auxiliaries)
is determined on the basis of load research survey. Based on a forecast of
the floor space, the energy sales forecast is developed.
The end use approach attempts to capture the impact of energy usage patterns
of various devices and systems. This method focuses on the end use of electricity.
The energy usage by various appliances and devices is calculated using the
E= S * N * P * H
E = Energy Consumption of an appliance (kWh),
S = Number of appliances per consumer,
N = Number of consumers,
P = Power requirement of the appliance (kW),
H = Hours of appliance usage (Hours).
The consumption of different end uses is summed up to arrive at the total consumption
in a particular sector. The appliance ownership and usage is determined for
different categories and classes of consumers. Advantage of this method is
that the socio-economic factors can be taken in to account.
Method 3 : Time Series Analysis
This method of forecasting future values of a variable involves the fitting
of a trend line to the historical data of a certain variable, using a method
of least squares. The method uses auto regression i.e. previous values of a
variable are used to predict the future values. The function used is as follows,
Yt = is the value of the variable under study,
t = time period, and
(Y(t-1),Y( t-2), Y(t-3) , …….. Y(t-k) ) are previous values of
The advantage of this method is its simplicity and ease of collection data
as the historical data for the dependent variable only, is required. The disadvantage
is that method does not describe the cause and effect relationship. Thus the
time series does not provide an insight in to the changes occurring in the
variable under study. This method is suitable for short term forecasts.
Method 4 : Trend Analysis
This method falls under the category of the non-causal models of demand forecasting
that do not explain how the values of the variable being projected and determined.
Here, we express the variable to be predicted purely as a function of time,
rather than by relating it to other economic, demographic, policy and technology
variables. This function of time is obtained as the function that best explains
the available data, and is observed to be most suitable for short term projections.
Forecasting of demand using this method is popular due to the ease of use and
simplicity of the method. Historical data for the variable under study is collected
and the same pattern is assumed to continue in future. The disadvantage of
this method is that economic and demographic, energy efficient technology factors
are not taken in to account.
STEP 3: Modelling the Demand Forecast
In modelling the demand forecasts, utilities can adopt a two way approach such
as energy sales approach as well as generation performance approach. The
generation performance approach considers historic unit performance, generation
capacity expansion, availability / plant load factor projections etc, it
captures only the restricted demand (constrained demand forecast) as well
as it dose not give insight of category / end use specific demand. In short
the generation performance approach does not quantify the unmet demand due
to the load shedding owing to the energy shortage situation within the region.
However, energy sales approach could be modified based on load shedding details
to take care of unmet demand and to get the unrestricted demand (unconstrained
demand forecast) in the region. The energy sales approach is also flexible
enough to incorporate the category specific / end use process demand and
its sensitivity to the diffusion of energy efficient technologies and DSM
The output of any load forecast exercise of most of the utilities includes
a forecast of the annual energy sales (in kWh) and the annual peak demand (in
KVA or kW). During the forecasting exercise it is worthwhile to forecast the
annual energy sales first and then use the forecasted energy sales in determining
the annual peak demand forecast after analysing the annual load factor. Since
the annual energy sales data is the integration of the hourly loads during
the year, it is less prone to weather and other spurious effects. A category
/ end use process specific forecasting method could be adopted by utilities
to forecast the energy consumption. Also, a combination of different forecasting
techniques could be used to suit the availability of data and information.
The profile of annual and daily demands could be determined based on historic
data. The load demands at other periods (seasonal variations and minimum loads)
shall be derived based on the annual peak demand and past pattern of load variations.
The best way of modelling the demand from DSM perspective, is based on scenario
analysis which compares different combinations of technological option to provide
the same level of energy services. The different scenarios could be constructed
and differentiated according to the level of projected energy service growth
(i.e. high, medium, low) and the degree of implementation of energy efficiency
improvements like diffusion of energy efficient technologies, DSM options etc.
The starting point to estimate the potential for various DSM options is to
construct at least two end use scenarios, a baseline scenario (scenario without
diffusion of energy efficient technologies and DSM options) and an energy efficient
scenario which considers improvements in end use efficiencies and possible
DSM options. However one can include some efficiency improvements in baseline
scenario as well based on the assumption that some efficiency improvement will
happen naturally without any specific market intervention to stimulate such
improvements. The scenario which dose not considers the improvement in end
use energy efficiency at all and thus maintaining the current levels of energy
efficiency is called as “Frozen Efficiency Scenario”.
The energy efficient scenario can be derived from the set of efficiency improvements
/ DSM options in several end uses and sectors or for one end use measure.
This efficient scenario can consider the possible technical improvements
processes and buildings that can be introduced in the projected year. However
potential for DSM / energy efficiency improvement can be characterised in
different ways, one way is to estimate the hypothetical savings that could
if all the systems could be converted instantaneously, whereas other way
is to estimate the savings that could be achieved if retiring systems were
with more efficient one. To asses the diffusion of energy efficient technologies
in place of retiring system various technology forecasting tools could be
In modelling the demand forecast, it is always advisable that discrete scenarios
such as high growth, moderate growth and low growth must be constructed
to examine the impact of variation of the key drivers and to assess the
range of demand
STEP 4: Reviewing the Results
In electric utilities there are a number of factors that drive the forecast
and the forecast further drives the various plans and decisions on investment,
construction and conservation.
With the onset of inflation and rapidly rising energy prices, emergence
of alternative fuels and efficient technologies (in energy supply and
end-use), changes in lifestyles, institutional changes etc. To deal with
all of the
above many forecasting techniques have been developed, ranging from very
methods to more complex time series techniques, for purposes of prediction,
which capture the effect of factors such as prices, income, population,
and other economic, demographic, policy and technological variables.
Though forecasting tools are available to aid in the process, the key
contributors in the process are the planning personnel with their knowledge
electricity consumers and an understanding of the way they use electricity.
for electricity is dependent on the magnitude and growth of the economy.
Robust economic growth creates more jobs, which leads to increased
population in service
territory, which, in turn leads to consumers who use more electricity.
Hence, demand forecast is not a one time exercise but needs to be constantly
against actual and updated for any major development or change in other
external demand drivers like policies, industry growth, changes in