|
APPLICATIONS OF NEURAL NETWORKS TO POWER SYSTEM CONTROL
Michal Kolcun - Radoslav Benč - Peter Szathmáry
The authors propose an overview of their research in the area of
artificial neural network (ANN) applications to power system
control. Application possibilities of ANNs are discussed. Then,
a description of three projects carried out at the Department of
Power Engineering FEI, Technical University in Kosice is given.
The first part of the presentation is a brief description of an
ANN based tool for fast, on-line contingency analysis. In this
tool, the ANN acts as a classifier, which detects possible
unsafe consequences of single line outage in power system. Two
different approaches to the problem are described and test
results are shown.
The second approach is an application of ANN to load flow
calculation. Load flow studies are important in power system
planning, control and operation. After training, the ANN is able
to approximate bus voltages when active and reactive loads are
presented. The approach was tested on a simple power system
model.
In the third part, a short term load forecaster is described. In
this approach, an ANN is the main component of the forecaster.
Choice of training data, normalisation, and design of the neural
network are discussed. The load forecaster performed load
forecasts for the Slovak Republic for the year 1996. Simulation
results, practical experiences, as well as possible improvements
are shown.
Keywords: power system control, neural networks, contingency analysis, load flow, load forecasting, backpropagation
|
