基于网络的自适应模糊推理系统在冰情预报中的应用

Appication of Adaptive-Network-Based Fuzzy Inference System to ice condition forecast

本研究将基于网络的自适应模糊推理系统应用在冰情预报中。通过分析基于网络的自适应模糊推理系统的网络结构、水温数据及其相关预报因子的分布特点、隶属度函数个数及预见期对预报结果影响的比较,确定了隶属度函数类型、隶属度函数个数和预报的预见期。文中以黄河宁蒙河段石嘴山为例,介绍水温预报的应用研究,并将预报水温同实测水温进行比较。通过描述预报值和实测值关系的确定性系数,对预报结果作了分析。结果表明,除了石嘴山水文站2002年预报结果实测值和预报值偏差较大、确定性系数偏小之外,其余预报组次中预报值和实测值均吻合较好,达到甲等预报方案。

The Adaptive-Network-Based Fuzzy Inference System （ANFIS） has been applied to forecast the ice condition. By using a hybrid learning procedure, the ANFIS is employed to model the nonlinear functions, such as the forecast. By analyzing the ANFIS structure, the characteristic of the water temperature data and the factors related to forecast, and comparing the forecasted result based on changing their numbers and days, the type and number of membership functions and the forecast period are selected. The above study is applied to forecast the water temperature of the Shizuishan Hydrological Stations in the Yellow River for 4-year winter. The forecast results are in good agreement with the observed data.