Monthly Runoff Forecasting Model Based on Electrostatic DischargeAlgorithm-Mixed Kernel SVM and Its Application

Monthly Runoff Forecasting Model Based on Electrostatic DischargeAlgorithm-Mixed Kernel SVM and Its Application

In order to improve the accuracy of hydrological forecasting, this paper establishes a support vector machine (SVM) based on the mixture of polynomial kernel and Gauss kernel, optimizes the key parameters and the mixed weight coefficients of mixed kernel SVM by the electrostatic discharge algorithm...

Full description

Saved in:
Bibliographic Details
Main Author: LI Xiangrong
Format: Article
Language:zho
Published: Editorial Office of Pearl River 2020-01-01
Series:Renmin Zhujiang
Subjects:
runoff forecasting
electrostatic discharge algorithm (ESDA)
mixed kernel function
Support Vector Machine (SVM)
parameter optimization
Online Access:http://www.renminzhujiang.cn/thesisDetails#10.3969/j.issn.1001-9235.2020.01.005
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:In order to improve the accuracy of hydrological forecasting, this paper establishes a support vector machine (SVM) based on the mixture of polynomial kernel and Gauss kernel, optimizes the key parameters and the mixed weight coefficients of mixed kernel SVM by the electrostatic discharge algorithm (ESDA), proposes a monthly runoff forecasting model in dry season based on mixed kernel ESDA-SVM, builds the Gauss kernel ESDA-SVM, polynomial kernel ESDA-SVM and ESDA-BP, as contrast forecasting models, as well as conducts the forecasting through the models with the data of the first 24 years and the last 10 years, taking the monthly runoff forecasting of a hydrological station in Yunnan from January to April in dry season as an example. The results show that the absolute average relative errors of forecasting by the mixed kernel ESDA-SVM model for monthly runoff from January to April are 4.09%, 3.32%, 3.51% and 5.64%, respectively. The forecasting accuracy of the mixed kernel ESDA-SVM model is higher than that of the other 3 models such as polynomial ESDA-SVM model. The mixed kernel ESDA-SVM model combines the advantages of polynomial global kernel function and Gauss local kernel function, so it is superior to the contrast models in forecasting accuracy and generalization ability, and has good practical application value.
ISSN:1001-9235

Similar Items