期刊
EXPERT SYSTEMS WITH APPLICATIONS
卷 198, 期 -, 页码 -出版社
PERGAMON-ELSEVIER SCIENCE LTD
DOI: 10.1016/j.eswa.2022.116934
关键词
Interval forecasting; Multi -objective optimization; Knee point; Hybrid kernel Rvm; Feature extraction
类别
资金
- National Natural Science Foundation of China [71671029]
- Major Program of National Fund of Phi-losophy and Social Science of China [17ZDA093]
This study aims to establish an integrated interval forecasting system for solar radiation, using feature extraction and a hybrid kernel relevance vector machine. The proposed system achieves higher coverage rate and narrower interval width in solar radiation forecasting.
In the context of global carbon neutrality, solar energy deserves more attention as a clean energy source. Ac-curate solar radiation forecasting techniques can provide favorable theoretical support for the siting and man-agement of solar power plants. This study aims to establish an integrated interval forecasting system to obtain better solar radiation interval forecasting results. In the proposed system, a feature extraction method based on mutual information is used to reduce the dimension of correlation variables. Then, a dual-channel input structure is constructed with the dimensionally reduced variables and lagged terms of solar radiation. A hybrid kernel relevance vector machine is developed to construct the forecasting intervals of solar radiation. In the hybrid kernel, a newly developed knee-based multi-objective jellyfish search is used to determine the weight of different kernel functions. According to the experiment results based on three American solar sites, it is demonstrated that the proposed system can obtain superior solar radiation forecasting intervals which provide higher interval coverage rate and narrower interval width. Through the optimization of the weight of the kernel function, we can obtain a forecasting interval that is more balanced between the forecasting coverage and the width of the forecasting intervals.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据