Tidal Current Short-Term Prediction Based on Support Vector Regression

2017 2nd Asia Conference on Power and Electrical Engineering

能源学;电工学

Yang, Guozhen^1 ; Wang, Haifeng^2 ; Qian, Hui^3 ; Fang, Jianming^3

Institute of Electrical Engineering, Chinese Academy of Sciences, University of Chinese Academy of Sciences, Beijing

100190, China^1

Institute of Electrical Engineering, Chinese Academy of Sciences, Beijing

100190, China^2

Zhejiang Chunan Stage Grid Corporation of China, Zhejiang

311700, China^3

关键词: Cross validation;    Harmonic method;    Model parameters;    Prediction model;    Short term prediction;    Support vector regression (SVR);    Support vector regressor;    Tidal currents;   

Others  :  https://iopscience.iop.org/article/10.1088/1757-899X/199/1/012024/pdf
DOI  :  10.1088/1757-899X/199/1/012024

来源: IOP

【 摘 要 】

The traditional method of short-term tidal current prediction, harmonic method, typically needs more than 18 years of history records. The method in the article uses univariate feature selection and F-test to reduce the dimension of the data fed to support vector regressor, which reduces the need of history records to less than a year. Model parameters are selected by grid searching and cross-validation. History records from two datasets are used to build prediction models, spanning 3 months and 1 year respectively. Mean average errors of both datasets after normalizing are less than 0.05.