Application of Neuro-fuzzy Model for Forecasting Water Quality (Chlorine) in Drinking Water Distribution Systems

Disinfecting water with chlorine is crucial for stopping the spread of infectious illnesses. The managers of water quality system would benefit significantly by predicting the free residual chlorine at major points of the water distribution system (WDS), since this would allow them to better assure the satisfaction and safety of their customers. Using a neuro-fuzzy model, this study makes an effort to forecast the chlorine concentration at various nodal points. Inputs to this model are Initial chlorine content, pH, Turbidity, Conductivity at the water treatment plant and output from the model is chlorine content at any location. Five hundred fifteen data were collected at regular interval at various stations as input and fed to the model for training. In this present work the used neural network is a four layers back propagation with multilayer perception system. Output of the Artificial Neural Network (ANN) is first estimated chlorine content which is fed to fuzzy controller along with other data. The result of the fuzzy controller is free residual chlorine contained at any point. Simulation results have been verified with the actual data. The performance evaluation was made and found to be more than 90%. The results obtained in the selected stations ware compared with the chlorine concentration to our developed neuro-fuzzy model. With using our newly developed model the chlorine concentration can be accurately predicated at a known location under any treatment plant.