07. Research and application of artificial neural networks to forecast and correct water level at rivers unaffected by tides

Abstract

Data of water flow and water level are very important in socio-economic development and national security. However, such data may be insufficient or discontinuously collected over a long period of time; thus, in some cases, data estimation is required. This paper proposes to use Multi Layer Perceptron - MLP neural network to model and calculate, estimate water level of rivers unaffected by tides with an allowed error. Based on actual monitoring data of a river's water level, MLP neural network was used to forecast river water level after one day and after ten days. Forecast results were compared with actual observed results in order to evaluate errors. Using MLP neural network to model and calculate, estimated river water levels were acceptable with the allowed error. Using this forecasting method, technicians should base on specific monitoring data and forecasting needs to adjust input and output parameters of the MLP network appropriately. Thus, this method could be applied to forecast river water level data according to actual requirements and support for the correction of hydrological data.