改进可变模糊ISODATA模型在径流预测中的应用  

Application of improved variable fuzzy ISODATA model in runoff prediction

作　　者：刘帅[1] 袁汝华[1] 

机构地区：[1]河海大学商学院,南京211100

出　　处：《信息技术》2016年第10期61-66,共6页Information Technology

基　　金：国家自然科学基金资助项目(40806018);国家水利公益性基金资助项目(201301055)

摘　　要：在可变模糊ISODATA预测模型的基础上,引入最佳模糊集划分理论,提出了可变模糊ISODATA与最佳模糊聚类有机结合的预测模型。其基本思路是从年降水量、年平均气温、年平均相对湿度、相对日照时数和年蒸发量5个方面构建预报因子,利用拉格朗日约束函数确定传统可变模糊ISODATA模型的最佳聚类数,通过对变系数模型参数滤定和年径流量数值模拟,求取符合精度要求的类别变量特征值与预测对象之间的回归方程进行预测。在沱沱河站年径流量预报中,预测效果比传统模糊ISODATA模型进行预测的效果要好,在短期3a内年径流量模拟相对误差均在10%之内,确定性系数达到0.8。该方法拓宽了可变模糊ISODATA理论在影响因素不稳定条件下年径流量研究的应用范围,为径流量的短期科学预测提供了一种新方法。A new prediction model using the combination of variable fuzzy ISODATA model is proposed by introducing best fuzzy set partition theory into the model using the combination of variable fuzzy ISODATA.Firstly,the based on the feature of runoff,from annual precipitation,annual mean temperature,annual average relative humidity,the number and annual evaporation it builds predictors. Secondly,use of Lagrange constraint functions improves the traditional variable fuzzy ISODATA the optimal number of cluster model. Finally,by using the coefficient of variation for model parameters and validation of annual runoff simulation,the regression equation of classified variable eigenvalues and prediction objective is solved for prediction. The results show that prediction accuracy of this new model than traditional fuzzy ISODATA model for predicting the annual runoff at Tuotuo river station,the relative errors of annual runoff simulation are less than 10%,and the uncertainty coefficient is 0. 8 or more in 3a,and this method broadens the scope of application of the theory of variable fuzzy ISODATA factors in predicting annual runoff study with unstable conditions,and it provides a new approach on scientific forecasting runoff.

关 键 词：可变模糊ISODATA 最佳模糊集 相对隶属度 约束函数 径流预测 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TV697.1[自动化与计算机技术—控制科学与工程]