The Application Of Probabilistic Netural Networks On Water Quality Assessment Of The Water Area Close To Three Georges Dam

The Three Georges Project is the key jproject for Changjiang river development. The impacts of project construction on ecological and environmental were always the concern of domestic and foreign researchers. The water environment status close to Three Georges Dam was investigated and studied using probabilistic neural networks method.The probabilistic neural networks model (PNN) for water quality assessment was developed in this thesis. For avoiding error occour in identifying water quality class with Gaussian function, a parameter which reduce the error in calculating Euclidean distance of the model was added into the factor of approaching centre from negative direction, and PNN water quality assessment model was improved (RPNN). Reliability , rationality and practicability of the model was verified by comparing analysis. The water environment status close to Three Georges Dam before and after impounding water was evaluated correctly. In addition, scientific and technical supports were supplied for water quality protection and supervision of The Three Georges reservoir.The water quality assessment on comprehensive factor, oxygen balance factor, nutrients factor and toxic factor in the Three Georges area was well generally and was not getting poor obviously after impounding water using RPNN model. Variation tendency of water quality during months was similar with that of before impounding water. Eutrophication in some local water area after impounding water may induce algae blooms. DO level in downstream of the dam was oversaturated.