Real-Time Prediction of Water Stage with Artificial Neural Network Approach

Abstract

An accurate water stage prediction allows the pertinent authority to issue a forewarning of the impending flood and to implement early evacuation measures when required. Existing methods including rainfall-runoff modeling or statistical techniques entail exogenous input together with a number of assumptions. In this paper, neural networks are used to predict real-time water levels in Shing Mun River of Hong Kong with different lead times on the basis of the upstream gauging stations or stage/time history at the specific station. The network is trained by using two different algorithms. It is demonstrated that the artificial neural network approach, which is able to provide model-free estimates in deducing the output from the input, is an appropriate forewarning tool. It is shown from the training and verification simulation that the water stage prediction results are highly accurate and are obtained in very short computational time. Both these two factors are important in water resources management. Besides, sensitivity analysis is carried out to evaluate the most suitable network characteristics including number of input neurons, number of hidden layers, number of neurons in hidden layer, number of output neurons, learning rate, momentum factor, activation function, number of training epoch, termination criterion, etc. under this specific circumstance. The findings lead to the reduction of any redundant data collection as well as the accomplishment of cost-effectiveness.