forecasting

Load

future needs of expansion ,equipment purchases

,or staff

Abstract-This work incorporates the different classification

of load forecasting which are being used by the power
industry (generation, transmission, distribution, retail), and
why load forecasting is very crucial for any operating and
planning utility and what are different techniques used to
perform load forecasting of a power grid.

hiring. LTLF the prediction time can be as long as

10years and above. A precise long term loadforecasting is essential for monitoring and
controlling power system operation.

1.Introduction
Load forecasting is very crucial to any industry.Load
forecasting means to predict the demand of load in advance.
Airline companies forecast the number of passengers to
schedule flights. Breweries forecast the beer consumption
to plan production. Retailers forecast the fashion goods
demand to decide what discount to offer to customers. Bank
forecast mortgage transactions to commit workforce in
branch offices [1].Results obtained from load forecasting
processes are used in different areas such as planning and
operation.
Planning of future investment for the construction depends
upon considerably on the accuracy of load forecasting [2].
Inaccurate load forecasts may lead to financial burden or
bankruptcy of a utility company .While load forecasting
provides a key input to power system operations and
planning, inaccurate load forecasts can lead to equipment
failures or even system-wide blackouts [1].
1.1 Classification of Load Forecasting
There is no single forecast that satisfy all of the needs of
utility. Therefore different forecasts are used for different
purposes. With so many applications ,it is unrealistic to
establish a forecasting problem for each application. The
classification of various forecasts not only depends upon
business needs of utilities, but also on other factors that
drive the electricity consumption .Based on the forecasting
horizon, we can classify load forecasting problems into the
following four groups [3]

Very
short
term
load
forecasting(VSTLF),forecasting horizon ranging
from a few ahead to few hours ahead.[4]
Short term load forecasting (STLF), one day to
two weeks ahead forecasting [4] used to supply
necessary information for the system management
of
day-to-day
operation
and
unit
commitment.STLF the prediction time is every
next hour, day by day, week by week and monthly.
Medium term load forecasting (MTLF) ,two
weeks to three years ahead forecasting, used for
the purpose of scheduling fuel supplies and unit
maintenance.MTLF the prediction time is 2-5
years.
Long term load forecasting(LTLF),three to fifty
years ahead forecasting[4],used to supply electric
utility company management with prediction of

2.Load forecasting Overview

2.1Characterstics of the Power System Load
The system load is the sum of all consumers load
at the same time. The objective of system STLF is
to forecast the future system load. Good
understanding of the system characteristics helps
to design reasonable forecasting models in
different situations. Various factors influence the
system load behavior ,which can be mainly
classified into the following categories:

Weather

Time

Economy

Random disturbances
2.2Characterstics of Developed Load forecasting
Methods:
In terms of lead time, load forecasting is divided
into four categories

Long-term forecasting with the lead time of more

than one year

Mid-term forecasting with the lead time of one

week to one year

Short-term load forecasting with the lead time of

1-168 hours

Very short-term load forecasting with the lead time

shorter than one day
3. Types of Load forecasting
3.1Residential Load forecasting
Residential customers include private households
that
use
energy
for
heating,cooling,cooking,lighting ,small appliances.
Among all customers classes, residential
customers have the most weather responsive
electricity consumption ,a residential load
forecasting model typically includes many

variables such
population.

as

weather

,calendar

and

many of good results were due to peeking the

future.[7]

3.2Commercial Load forecasting

4.1.4Univariate Models

Commercial customers are those customers not

involved in manufacturing such as retail
stores,restaurants,hotels
,and
educational
institutions. Commercial customers can be roughly
divided into two sub-classes small commercial and
large commercial customers, load forecasting for
small commercial customers is similar to
residential .Load forecasting for large commercial
customers requires customized efforts depending
upon the type of business. Some of them have
strong seasonal patterns for instance the load of
hotels in vacation area is mainly affected by
tourism demand during holiday seasons.

Exponential smoothing assigns exponentially

decreasing weights to the past observation over the
time. It does not rely on variables other than
lagged loads ,meaning less data requirements than
other widely used techniques such as MLR and
ANN.Since the electricity demand is highly driven
by weather ,changes in weather patterns can
greatly affect the load profiles. When the weather
condition is quite volatile ,the techniques that do
not leverage meteorological forecasts are often in a
disadvantage situation.

3.3Industrial Load forecasting

4.2.1Similar Day Method

Industrial load is very unpredictable. There are

several factors affecting the load of an industrial
customer ,such as facilities, business types
,production levels ,electricity rates, customerowned generation.

The similar day method is to find a day in the

history that is similar to the forecasted day. The
similarity is usually based on day of the week,
season of a year ,and weather patterns. The similar
day method is one of the earliest methods being
applied to load forecasting. Even today, many
system operators are still having the load and
temperature profiles of the representative days
hanging on the wall of operations room. Modern
similar day method is often implemented with
some clustering techniques. Instead of one similar
day ,the algorithm m ay identify several similar
days or similar segment of a day ,and then
combine them to obtain the forecasted load profile.

4.LoadForecasting
Methodologies

Techniques

and

4.1Techniques
4.1.1 Linear Regression Model
Regression analysis is a statistical process for
estimating the relationship among variables.
Regression models have been used for both
STLF and LTLF.
4.1.2Semi-parameter additive models
The semi-parametric additive model is in the
regression
framework[5],but
design
to
accommodate some non-linear relationship and
serially correlated errors. In particular ,such
models allow non-linear and non-parametric terms
using the framework of additive models. In load
forecasting ,these generalized additive models are
used to estimate the relationship between load and
explanatory variable temperature, calendar,[6].
4.1.3Artificial Neural Network
The ANN have been extensively used for load
forecasting since 1990s,ANN is a soft computing
technique that does not require the forecaster to
explicitly model the underlying physical system.
By simply learning the patterns from historical
data , a mapping between the input variables and
the electricity demand can be constructed and then
adapted for prediction. A number of ANN
architectures have been used for load forecasting
,such as back propagating Hopfield ,Boltzmann
machine, among which the most popular one is
back propagation. Researchers have been reporting
fairly good results with ANN models ,though

4.2Methodologies

4.2.2Variable Selection
For many techniques that rely on explanatory
variables ,an important step is to determine which
explanatory variables to use and their functional
forms .
4.2.3Weather Station Selection
Since weather is a major driving factor of
electricity demand, it is important to figure out the
right weather station for a territory of interest.
4.2.4Hierarchial Forecasting
Due to deployment of smart grid technologies
,how to utilize the hierarchies to improve load
forecasts becomes an important topic in todays
load forecasting community.
5.Load Forecasting
Solutions

Software

Tools

and

There are many tools available in market that may

be used for energy forecasting purposes. They all
have different learning curves, levels of technical
support , depth of forecasting procedures , and
price tags. There is not yet one single tool that
dominates all metrics. When considering which
tool to use ,utilities have to evaluate many factors,
such as direct(i.e., license and service fees) and
indirect costs (salaries and training costs for the

user ) of the software package ,potential value

add ,and implementation time. This section
introduces several commercial load forecasting
tools and solutions that have been widely used in
the utility industry.

accuracy of load forecasting depends not only on

the load forecasting techniques, but also on the
accuracy of forecasted weather scenario.

5.1SAS

REFERENCES

SAS Energy Forecasting is built on the SAS

family of software products which have been used
by electric utilities since 1976.The solution is
tailored for electric utility energy and load
forecasting, automatically stepping as through as
many as nineteen models to select the best forecast
model. For inexperienced forecasts, the process
can be highly automated with few decisions
required ,while experienced users can expand the
models with additional variables or import models
into the solutions. The SAS Energy Forecasting is
transparent ,with model result at all stages of
process available for review and to archive for
regulatory documentation.

[1]Tao Hong Big Data

Laboratory (Big DEAL)

5.2Itron

[6]Hippert, H.S., Pedreira, C.E., & Souza. Neural

Network for Short Term Load Forecasting

MetrixND and MetrixLT are windows application

designed specifically for utility forecasting
processes. Some of the high level features are:

Support for annual, quarterly ,monthly, daily,

hourly and sub hourly data

Folder based project explorers to organize tables

,variables ,configured charts and configured
reports

Drag and drop functionality for building

transformations ,models, reports and other objects

Exponential smoothing
5.3Integral Analytics
LoadSEER(spatial Electric expansion & risk)is
spatial load forecasting software tool designed
specifically for transmission and distribution (T &
D)planners who face increasingly complex grid
decisions caused by emerging micro grid
technologies ,extreme weather events, and new
economic activity. The objective of LoadSEER is
to statistically represent the geographic
,economic , distributed resources, and weather
diversity across a utilitys service territory ,and use
that information to forecast circuit and bank level
peak loads ,sub sections of the circuit ,acre level
changes ,and impact from various scenarios over
the planning horizon.
6.CONCLUSION
Precise load forecasting is very important for
electric utilities in a competitive environment
created by the electric industry deregulation. The
development and improvement of appropriate
mathematical tools will lead to the development of
more accurate load forecasting techniques. The

Energy

Analytics

[2]R. Behera , B.B. Pati , B.P. Panigrahi A Long

Term Load Forecasting of an Indian Grid Of
Power System Planning
[3]Eugene A. Feinberg, Dora Genethliou State
University of New York, Stony Brook
[4]Heiko
Hahn
Silia
Meyer-Nieberg
Stefan ,Electric Load Forecasting Methods-Tools
for decision making
[5]Hyndman, R.J., and Fan,S.(2010).Density
forecasting for long term peak electricity demand

[7]Fan,S.,Methaprayoon, k.,&Lee,W.J.Short term

Load Forecasting based on a semi-parametric
additive model