WANG Tao, YANG Kai-lin, GUO Xin-lei, FU Hui (State Key Laboratory of Simulation and Regulation of Water Cycle in River Basin China Institute of Water Resources and Hydropower Research, Beijing 100038, China)

Abstract: Forecast on freeze-up water temperature is a basis for the ice condition. In this study, Adaptive-Network-based Fuzzy Inference System (ANFIS) and Artificial Neural Networks (ANN) are applied to the freeze-up water temperature forecasting in the Yellow River. Reasonable input parameters of the two models are determined through analyzing each model characteristic, information of water temperature and related factors. For better comparability, the same input factors and forecast periods are used to estimate the 4-year water temperatures in Bayangaole, Shanhuhekou and Toudaoguai hydrometric stations, which are located in the most north of the Yellow River. The forecast results are assessed by coefficients of determination, correlation coefficients and root mean square errors. A comparison of ANFIS and ANN results shows that ANFIS gave better results than ANN by 12 forecast cases. As a result, ANFIS model is founds to be superior to ANN model for forecasting the time series information, such as freeze-up water temperature.

Key words: ANFIS, ANN, forecast, ice condition, water temperature, Yellow River

Published in: Journal of Hydraulic Engineering, Vol. 44, No. 7, 2013