基于改进时间卷积网络与藤Copula的短期风速预测  

Short-Term Wind Speed Prediction Based on Improved Temporal Convolutional Network and Vine Copula

作　　者：黄宇[1] 张宗拾 刘家兴 李旭昕 张鹏[1] HUANG Yu;ZHANG Zongshi;LIU Jiaxing;LI Xuxin;ZHANG Peng(Department of Automation,North China Electric Power University,Baoding 071003,China)

机构地区：[1]华北电力大学自动化系,河北保定071003

出　　处：《电力科学与工程》2024年第7期60-69,共10页Electric Power Science and Engineering

基　　金：中央高校基本科研业务费专项资金资助项目(2021MS089)。

摘　　要：考虑风电场相邻风机风速间以及风速与气象因素间复杂的非线性关系,提出了一种基于改进时间卷积网络与藤Copula相结合的风速预测方法。首先,利用深度残差收缩网络中存在的注意力机制及软阈值化的思想改进时间卷积网络中的残差模块,并进行初步风速预测;然后,考虑到众多气象因素对风速的影响,使用核主成分分析对气象数据进行降维,在保证数据特征的同时,降低数据的复杂度;最后,利用藤Copula在描述非线性相关结构方面的优势构建修正模型,使用降维的气象数据修正初步风速预测值,得到最终的风速预测结果。实验证明,所提方法提高了短期风速预测的精度。A wind speed prediction method based on improved temporal convolutional network(TCN)combined with vine Copula is proposed considering the complex nonlinear relationships between wind speeds of adjacent wind turbines and between wind speeds and meteorological factors in wind farm.Firstly,the residual module in TCN is improved with the attention mechanism and soft-thresholding idea from deep residual shrinkage network(DRSN)for preliminary wind speed prediction.Then,acknowledging the influence of numerous meteorological factors on wind speed,kernel principal component analysis(KPCA)is used to reduce the dimensionality of meteorological data,and reduce the complexity of data while preserving data features.Finally,leveraging the advantages of vine Copula in describing nonlinear correlation structures,a correction model is constructed.By utilizing the dimensionally reduced meteorological data to correct the preliminary wind speed prediction values,the final wind speed prediction results are obtained.The experimental results show that the proposed method improves the accuracy of short-term wind speed prediction.

关 键 词：风速预测 改进时间卷积网络 气象因素 核主成分分析 藤Copula 

分 类 号：TM614[电气工程—电力系统及自动化]