Using soft computation to predict the effluent quality from municipal and industrial wastewater treatment plants

• Main Authors: Hsin-Hsien Ho, 何欣憲
• Other Authors: Tzu-Yi Pai
• Format: Others
• Language: zh-TW
• Published: 2006
• Online Access: http://ndltd.ncl.edu.tw/handle/vk2fu8

碩士 === 朝陽科技大學 === 環境工程與管理系碩士班 === 94 === The primary goals of the study are as follows: (1) to utilize grey relational analysis to select system parameters of wastewater treatment plants (WWTPs) and use backpropagation neural network (BNN) and fuzzy system with genetic algorithm (GA) to construct models for predicting effluent quality from WWTPs; (2) to predict effluent quality from WWTPs using grey model (GM) and verify long term prediction; and (3) to compare the effects of three different models for predicting the effluent quality from WWTPs. The results indicated that the consistency of on-site monitoring data when using GABNN and GA Fuzzy System was the best. The consistency of other effluent pollutants was not as good as that of on-site monitoring data. The consistency of effluence nutrients was the worst. The consistency of effluent pH was the best when using both models, that of effluent chemical oxygen demand was worse, and that of effluent suspended solids was the worst. In the aspect of GM (1, N) model, it revealed that the consistency between the was good at the commence of prediction, but it would diverse latter. It suggested that GM (1, N) was good for short-term prediction but not suitable for long-term prediction. When using rolls grey model (RGM (1,N)) to predict the effluent quality from an industrial WWTPs, the consistency was not good. But the consistency was good when using RGM (1, N) to predict the effluent quality from a municipal WWTPs. It suggested that RGM (1, N) was capable for long-term prediction.