基于SVMD-Informer-XGBoost的风电功率区间预测
Wind Power Power Interval Prediction Based on SVMD-Informer-XGBoost
黄丹 1张历卓 1陈思羽 1匡迎春1
作者信息
- 1. 湖南农业大学信息与智能科学技术学院,湖南长沙 410128
- 折叠
摘要
为了提高风电功率区间预测的精度,文章提出逐次变分模态分解(SVMD)、Informer和XGBoost算法模型的结合,实现风电功率区间预测.首先,SVMD算法自适应地分解出风电功率数据的各模态分量(IMF)和残差分量.其次,通过网格搜索算法(CV)对Informer和XG-Boost 模型算法进行优化,并通过网格搜索算法确定Informer-XGBoost输出预测值的最优权重,构建基于SVMD-Inform-er-XGBoost 的预测模型.最后,通过分位数回归法(Quantile regression)输出预测值和真实值的误差,得到风电功率的区间预测.实验结果表明,所提出的方法相较于传统单一模型,在区间预测的准确度和稳定性上有了显著提升.在9 5％置信度水平下,区间预测指标区间覆盖率(PICP)和区间平均宽度(PINAW)分别可达 94.96％和 0.0362.
Abstract
In order to improve the accuracy of wind power interval prediction,this paper propo-ses a combination of successive variational mode decomposition(SVMD),Informer and XG-Boost algorithm models to achieve wind power interval prediction.First,the SVMD algorithm adaptively decomposes each modal component(IMF)and residual component of wind power da-ta.Secondly,the Informer and XGBoost model algorithms are optimized through the grid search algorithm(CV),and the optimal weight of the Informer-XGBoost output prediction value is de-termined through the grid search algorithm to build a prediction model based on SVMD-In-former-XGBoost.Finally,the quantile regression method is used to output the error between the predicted value and the true value to obtain the interval prediction of wind power.Experimental results show that the proposed method has significantly improved the accuracy and stability of interval prediction compared with the traditional single model.At the 95％confidence level,the interval prediction index interval coverage(PICP)and interval average width(PINAW)can reach 94.96％and 0.0362 respectively.
关键词
逐次变分模态分解/XGBoost/Informer/网格搜索Key words
Successive variational mode decomposition/XGBoost/In former/Grid search引用本文复制引用
出版年
2024
NETL
NSTL
万方数据