Forecasting chlorine residuals in a water distribution system using a general regression neural network

In a water distribution system (WDS), chlorine disinfection is important in preventing the spread of waterborne diseases. By strictly controlling residual chlorine throughout the WDS, water quality managers can ensure the satisfaction and safety of their customers. However, due to the travel time of water between the chlorine dosing point and any strategic monitoring points, water treatment plant (WTP) operators often receive information too late for their responses to be effective. Given the ability to forecast the chlorine residual at strategic points in a WDS, it would be possible to have superior control over the chlorine dose, thereby preventing incidents of underand over-chlorination. In this research, a general regression neural network (GRNN) has been developed for forecasting chlorine residuals in the Myponga WDS to the south of Adelaide, South Australia, 24 hours in advance. A number of critical model issues are addressed including: selection of an appropriate forecasting horizon; division of the available data into subsets for modelling; and, the determination of the inputs that are relevant to the chlorine forecasts. In order to determine if the GRNN is able to capture any nonlinear relationships that may be present in the data set, a comparison is made between the GRNN model and a multiple linear regression (MLR) model. When tested on an independent validation set of data, the GRNN models were able to forecast chlorine levels to a high level of accuracy, up to 24 hours in advance. The GRNN also significantly outperformed the MLR model, thereby providing evidence for the existence of nonlinear relationships in the data set.