基于互信息的湖泊日水位预测——以西洞庭湖为例

发布时间：2018-05-11 06:13

  本文选题：互信息 + 湖泊水位预测　； 参考：《人民长江》2017年16期

【摘要】：鉴于传统的湖泊水位预测在输入因子选择时具有一定的盲目性,以西洞庭湖为例,利用基于互信息的输入因子选择法建立了日水位预测模型。按河流生态功能将水文年划分为枯水期、汛前涨水期、汛期、汛后退水期4个时期,然后分期计算影响湖泊日水位的自变量与日水位的互信息,并引入广义相关系数将互信息归一化,选出各时期互信息最大的自变量因子作为模型的输入变量。经过模型计算与数据分析可得:F检验结果显著,回归值与实测值的相关度高,剩余标准差小。由此证明用互信息筛选出的因子作为模型的输入变量能取得较好的精度并在实际中易于操作。
[Abstract]:In view of the blindness of the traditional prediction of lake water level in the selection of input factors, an input factor selection method based on mutual information is used to establish a daily prediction model of water level in the west of Dongting Lake. According to the ecological function of rivers, the hydrological year is divided into four periods: dry season, flood period before flood and receding period, and then the mutual information between independent variables and daily water level affecting the daily water level of lakes is calculated by stages. Then the generalized correlation coefficient is introduced to normalize the mutual information, and the independent variable factor with the largest mutual information in each period is selected as the input variable of the model. Through the model calculation and data analysis, we can find that the result of the test is remarkable, the correlation between the regression value and the measured value is high, and the residual standard deviation is small. It is proved that the factors filtered out by mutual information as input variables of the model can obtain good accuracy and are easy to operate in practice.
【作者单位】： 武汉大学水资源与水电工程科学国家重点实验室;武汉大学水资源安全保障湖北省协同创新中心;
【基金】：国家自然科学基金项目(51179130) 国家重点研发计划课题(2016YFC0401306)
【分类号】：P338
，

本文编号：1872758