Stimulation Initiative for European Neural Applications
Esprit Project 9811
Monitoring of Water Dam
Company background
Tauernkrafwerke AG, located in Salzburg, is Austrias largest provider of
hydro dam generated electrical energy.
The problem
A water dam is a subject to deformation due to various influences
inculding the height of the water level and temperature. In order to
assess the security of the dam, it is important to model the
displacement of the dam and to monitor the deformation by comparing
measured data to the model. In case of significant differences between
measured and modelled displacement, an investigation of the dam has to
be carried out.
Neural network application
The objective of this project has been to develop a model to predict the
deformation on the basis of a neural network. Influencing factors in the
modelling of the dam-displacement such as the time of measurement ,
water-level, and air temperature have been taken into consideration.
Benefits
The neural network method provides a significant improvement in the
forecast compared to conventional statistic approaches (e.g. multiple
regression). Depending on the forecast period, the absolute error for
the neural network forecast model lies between 0.1 % and 1.0%. In
comparison, the conventional multiple regression based model achieves
absolute errors of 2% - 3%. The neural network method supplies a precise
forecast which reduces unnecessary investigations of the dam and thereby
reduces operational costs.
Generalization
The feature of this application is the use of neural networks for
forecasting purposes in fields where conventional statistic approaches
achieve only insufficient results.
Contact person
Tauernkraft AG -
Walter Köhler.
Postfach 161,
Rainerstraße 29,
A-5021 Salzburg,
Austria.
Fax: +43 662 88 950 63.