Full Description

Deterministic models have been widelyused to predict water quality in distributionsystems, but their calibration requires extensiveand accurate data sets for numerousparameters. In this study, alternative data-drivenmodeling approaches based on artificialneural networks (ANNs) were used topredict temporal variations of two importantcharacteristics of water quality, chlorineresidual and biomass concentrations.The authors considered three types of ANNalgorithms. Of these, the Levenberg-Marquardtalgorithm provided the best results inpredicting residual chlorine and biomass witherror-free and "noisy" data. The ANN modelsdeveloped here can generate water qualityscenarios of piped systems in real time to helputilities determine weak points of low chlorineresidual and high biomass concentration andselect optimum remedial strategies.

Includes 23 references, tables, figures.

Product Details

Edition: Vol. 102 - No. 7 Published: 07/01/2010