Przemysl Chemiczny, Vol.89, No.9, 1236-1240, 2010
Use of SVM neural networks to prediction of the cooling water temperature in the PKN Orlen rafinery
Hydraulic and thermal models of an industrial cooling water supply system were developed by using the support-vector-machine neural networks to predict temps. of the cooling water in 96 points of the waterflow network. The network consisted of 1614 junctions, 9 tanks, 1188 pipelines, 26 pumps and 640 valves. A good agreement of the predicted and measured temps. was achieved.