Comparative Study on River Flow Forecasting Methods of River Networks

Abstract

This paper attempts to set up multivariate linear regression analysis (MLRA) model and 3-layers BP artificial neural network (ANN) mode on river networks and do some comparative researches about them. The applications to the watershed of Tarim indicate that the river flow processes which are simulated separately by two models are satisfactory. They can be the foundation for water resource allocation and scheduling. Above all, through analyzing the structures and forecast precisions of these models, artificial neural network model is better as compared with multivariate linear regression analysis model. In the end, this article puts forward some proposals about how to strengthen the predict abilities of river flow forecasting methods of river networks.