Using Neuro-Fuzzy To Forecast Ice Condition

The ice conditions are natural phenomena appeared in the rivers of cold regions, which have important impacts on hydraulic engineering. The disaster of the ice condition has increased in both the degree and range in the recent years, which damage to the economic and social development is more serious. Therefore, accurate forecast of the ice conditions is very important for against the potential disaster. Therefore, the Neuro-Fuzzy theory is applied to forecast ice conditon in this study,. The innovations are are follows.(1) An the artificial neural networks (ANN) model based on feed-forward back-propagation (FFBP) and improved by Levenberg-Marquardt (L-M) algorithm is developed to forecast the ice condition. Because of the complexity of ice conditions, traditional methods could hardly give accurate prediction in the ice condition forecast, while ANN have obvious advantage over other traditional methods for forecasting ice condition. An ANN model based on BP and improved by L-M algorithm is developed to forecast the ice condition including ice run date, freeze-up date, break-up date and so on in the Inner Mongolia Region of the Yellow River. The forecast results are in good agreement with the measured ones. Simulation also shows that the ANN model is superior to the MLR model and GM(0,1) model. This forecast system has been runing for the several years to provide the reliable information of Yellow River in the winter management.(2) Adaptive-Network-based Fuzzy Inference System(ANFIS) is applied to ice condition. With its hybrid learning scheme, ANFIS, constructed under the framework of the neural networks and fuzzy models, and the latter possess certain advantages over the former two, is convenient for modelling the nonlinear multivariable process. Consequently, ANFIS is applied to the freeze-up water temperature forecast in Bayangaole, Shanhuhekou, Shizuishan and Toudaoguai Hydrometric Station of Inner Mongolia reach of the Yellow River. Through such comparisons, it was discovered that the water temperature forecast results approximately agree with those of the field records. Compared with the results of ANFIS and ANN, ANFIS model was founds to be superior to ANN model for forecasting the time series information.(3) Chinese Calender is introduced to forecast the ice condition. In this study, the essence and characteristic are analyzed theoretically for Lunar Calendar, Solar Calendar and Chinese Calender. Though reasearch on the relationship between the Chinese Calender and modern hydrology, it is reasonable and scientific to apply the coordinate system of Chinese Calender to forecast hydrological information. Chinese Calender is associated with the ANN to forecast ice condition in South-to-North Water Diversion Middle Route Project. It is difficult to forecast ice condition because there are no existing hydrological data for this building Project. By analyzing the meteorological conditions, BP-ANNs model improved by L-M algorithm is applied to forecast the date of temperature downcrossing0℃and the date of temperature upcrossing0℃in the coordinate system of Chinese Calender and Solar Calendar, respectively. Obviously, the accuracy is more improved when the Beginning of Winter is as the datum points of the statistic date in the coordinate system of Chinese Calender than that in Solar Calendar. This result show that the Twenty-four Solar Terms is as an important reference for the hydrological forecast correlated the weather and Climate.