Abstract:
This thesis applies an Artificial Neural Network (ANN) to build a model for predicting residual chlorine levels at the Metropolitan Waterworks Authority Bangkhen (MWAB). This plant is so complex that a simple mathematical model can hardly be obtained resulting to difficulty in added chlorine level determination in treatment process. With the prior collected chemical uses data, thus ANN is a promising method to build a mathematical model of water treatment process. This process has several parameters such as water distribution flow rate, dosed chlorine and residual chlorine in water treatment process. These parameters can be classified by behavior of consumer and chlorine dosed by chemists. The network models were trained by using of a set of real data sampled every 2 hours in the year of 2011. Finally, a best chosen network model has structure of 10-18-1 feedforward network with the performance of 0.008 in mean sum of square error.
The validation and implementation of model at Bangkhen water quality analysis laboratory show that a accurate prediction of ± 0.2 ppm. This error is accepted by MWAB. In conclude, the designed ANN model makes a good deal for high quality water distribution, low decay rate of chlorine, high consumer satisfaction and good health.