The electric utility planning process begins with the
electric load forecasting, because of the advanced need
for new utility plants. These long lead times require the
utility planning horizon to be at least ten years long.
Since utility decisions involve an economic analysis of
the operating and investment costs, the utility planning
horizon may range from fifteen to thirty years into the
future. Forecasting load demand is4a difficult procedure
and combines art with science. The key contribution of
forecasters is their knowledge of electricity consumers
and an understanding of the way they use electricity and
other competing energy forms. The problem gains
special aspects in developing countries, such as Egypt,
because of the high demand growth rate as well as the
wide differences in the modes and levels of
consumption in the various regions (govemo rates) in
the country. During the recent years, some new
mathematical tools have been published such as expert
system (EXP.), Artificial Neural Network (ANN) and
Fuzzy logic systems. These tools almost replaced the
classic methods used by most utilities and research
centers personnel for forecasting. In this study, a
technique based on the Artificial Neural Network
(ANN) method, is used to estimate Peak load and Light
load for the Egyptian power system network as an
example for developing countries. This technique is
highlighted by the accuracy and sensitivity of the model
with respect to the ANN parameters. The proposed
technique can thus be applied to simple as well as
extended power system networks.
Consequently, in this study, several structures for
Neural Networks are proposed and tested. They
proved to perform as one of the best and most
sophisticated forecasting systems. In this study, the
case of a number of neurons layers equal 7, gives the
best results with high accuracy with the least error.
The forecasted Peak loads and Light loads, up to year
2010, for the six Regions of the Egyptian Unified
Network; Alexandria, Delta, Cairo, North Upper
Egypt, South Upper Egypt and the Cval, are obtained
directly from one case by using the actual and practical
past ten years data.
Key Words
Planning, Load Forecasting, Econometric
Method, Artificial Neural Network. |
