基于回归支持向量机的水库防洪承载力预测模型研究
Forecasting model of reservoir''s flood control capacity based on support vector machine for regression
投稿时间:2020-12-01  修订日期:2020-12-01
DOI:
中文关键词:  水库防洪承载力预报  回归支持向量机  贝叶斯优化  核函数
英文关键词:Reservoir flood control capacity prediction  regression support vector machine  Bayesian optimization  kernel function
基金项目:国家重点研发计划(2016YFC0402703);国家重点研发计划(2019YFC0409000)
作者单位邮编
王正华 河海大学水文水资源学院 210098
包为民 河海大学水文水资源学院 210098
孙逸群 河海大学水文水资源学院 
侯露 河海大学水文水资源学院 
摘要点击次数: 106
全文下载次数: 0
中文摘要:
      防洪承载力,即水库目前剩余防洪库容条件下,不泄洪所能容纳的流域面降雨量。根据水量平衡原理分析概化出防洪承载力的预报因子为当前土壤含水量,当前入库流量以及水库剩余库容。利用2010-2020年青山水库55场历史洪水建立基于回归支持向量机的防洪承载力预测模型,利用贝叶斯优化进行超参数率定,通过分析预测值确定回归支持向量机核函数为线性,预测值与实测值相关系数为0.9527,平均绝对百分误差为24.2853%,预测偏小百分比为49.0909%,表明模型精度较高且计算结果较为稳定,可为水库防洪提供参考。
英文摘要:
      Flood control capacity means the area rainfall of the basin that can be accommodated without discharge under the current remaining flood control capacity of the reservoir. According to the principle of water balance, the predictors of the flood control capacity are summarized as the current soil water content, the current inflow flow and the remaining capacity of the reservoir. Establish a flood control carrying capacity prediction model based on regression support vector machine using the data of 55 historical floods in Qingshan Reservoir from 2010 to 2020. Using Bayesian optimization to calibrate the hyperparameters, the kernel function of the regression support vector machine is determined to be linear by analyzing the predicted value. The correlation coefficient between the predicted value and the measured value is 0.9527, the average absolute percentage error is 24.2853%, and the under-predicted percentage is 49.0909%, indicating that the model has high accuracy and the calculation result is relatively stable, which can provide a reference for reservoir flood control.
  查看/发表评论  下载PDF阅读器