Analysis on Forecasting Runoff in Tarim River Basin
Tang De-shan · Water Conservancy Science and Technology and Economy · 2006
Due to the characteristics of randomicity and gray information in forecasting runoff,the traditional models were quite difficult to solve those complex and non-structural problems.Based on the runoff forecasting of some gauge stations in Tarim River Basin participated,the validity and some problems of GM(1,1) and BP algorithm existed in runoff forecasting were discussed.In the paper,the gray absolute,relative and integrated correlation degree between characteristic series and factors' series were firstly analyzed using gray correlation theory.And then,a GM(1,1) was established.The limitation of simulating a stochastic and oscillating sequence through GM(1,1) under the condition of the satisfaction of quasi-smoothness of original sequence and the satisfaction of quasi-exponential law of 1-Accumulating Generation Operator was talked over.In the end,an artificial neural network model was set up.The weight matrices and biases of the network were trained using BP algorithm.The historical evolvement of runoff was simulated and the runoff of reserved years were predicted and checked up.