STATISTICAL LOAD MODELING
Eugene A. Feinberg,
Department of Applied Mathematics and Statistics
State University of New York at Stony Brook
Stony Brook, NY 11794-3600, USA
Janos T. Hajagos
Long Island Power Authority (LIPA)
175 East Old Country Road
Hicksville, NY 11801, USA
This paper discusses the improvement of the statistical
model that was developed in  by adding a new weather
variable called sunshine.
We also took into account other
weather factors such as ambient temperature, humidity,
wind speed, and sky cover as well as time factors such as
the day of the week and the hour of the day.
developed a statistical model that describes the electric
power demand during a summer period for close
geographic areas that are called load pockets.
proposed method was evaluated on real data for several
load pockets in the Northeastern part of the USA.
Load Pocket Modeling, Load Forecasting
Accurate models for electric power load forecasting
are essential to the operation and planning of a utility
Load forecasting helps an electric utility in
making important decisions including decisions on
purchasing and generating electric power, load switching,
area planning and development.
Demand for electric
power typically depends on the temperature and several
other weather factors as well as the day of the week and
the hour of the day.
These factors are included in this
Load forecasts can be divided into three categories:
short term forecasts which are usually from one hour to a
week, medium forecasts which are usually from a month
to a year or even up to three years, and long term
forecasts which are over three years .
mathematical, econometric, and other load models.
In this paper we describe an improved statistical model
for a load pocket compared to our model that was
developed in .
This model takes into account weather
parameters, a day of the week, and an hour during the
In  four different weather parameters have been
considered: the temperature, humidity, sky cover, and
In this paper we include another variable
This new variable indicates when the sun
is up. We developed an algorithm to compute the model
parameters and tested the importance of particular
The algorithm finds model
parameters by performing a sequence of linear
The model parameters were estimated using
two different approaches.
The model includes weather
parameters as well as time parameters that consist of a
day of the week and an hour during the day.
advantages of this model are its accuracy, simplicity, and
the use of only two types of data: weather and load.
In the literature there are several papers discussing