The development of an artificial neural network (ANN) model for river flow forecasting for theUpper and Lower Humber River Basin will be presented in this paper. Two types of ANN wereconsidered, general regression neural network (GRNN) and the back propagation neural network(BPNN). GRNN is a nonparametric method with no training parameters to be adjusted during the trainingprocess. BPNN on the other hand has several parameters such as the learning rate, momentum, andcalibration interval, which can be adjusted during the training to improve the model. Since differentcombination of parameters used in the BPNN produced only slight differences in the results, the defaultsetting of training parameters was used. One and two-day ahead forecasts were obtained from the twoANNs using precipitation, cumulative degree-days, and flow data all suitably lagged as inputs. It wasfound that the GRNN model produced better forecasts than the BPNN for the Upper Humber and bothmodels performed equally well for the Lower Humber . The ANN approach also produced much betterforecasts than the current routing method used by the Water Resources Management Division of theNewfoundland and Labrador Department of Environment and Conservation.
展开▼
