Abstract

This paper describes a systematic methodology based on artificial neural networks for model identification and its application to the prediction of water chemical properties under normal operation conditions in a power plant. The model obtained allows to detect incipient anomalies by comparison between the real and predicted values.