基于时空关联分解重构的风速超短期预测  

Multi-Step Ultra-Short-Term Wind Speed Prediction Based on Decomposition and Reconstruction of Time-Spatial Correlation

作　　者：潘超[1] 李润宇 蔡国伟 王典 张永会 Pan Chao;Li Runyu;Cai Guowei;Wang Dian;Zhang Yonghui(Key Laboratory of Modern Power System Simulation and Control&Renewable Energy Technology Ministry of Education,Northeast Electric Power University,Jilin,132012,China;Songhuajiang Hydropower Co.Ltd Jilin Baishan Power Plant,Jilin,132400,China)

机构地区：[1]现代电力系统仿真控制与绿色电能新技术教育部重点实验室(东北电力大学),吉林132012 [2]松花江水力发电有限公司吉林白山发电厂,吉林132400

出　　处：《电工技术学报》2021年第22期4739-4748,共10页Transactions of China Electrotechnical Society

基　　金：国家重点研发专项(2016YFB0900100);常规水电站结合抽蓄、光伏、风电、电化学储能联合开发研究(525687200009)资助项目。

摘　　要：考虑风速的空间关联性进行多步预测是规模化风电并网的研究热点,该文采用一种改进的多位置多步风速预测方法。首先,提出风速矩阵时空关联分解重构策略,对风场内各风机进行灰色关联分析,并据此利用时序控制的空间关联优化算法进行优选排序,获取典型风机及临近域空间信息,对该空间信息进行重构,以提高空间特征提取效率;然后,将重构的时空三维信息输入卷积记忆网络,以降低信息缺失对预测精度的影响,并进行空间特征提取及多步超短期预测;最后,通过对不同风电场的风速及风功率进行预测,验证所提方法的预测精度和泛化能力。Considering the spatial correlation of wind speed makes multi-step prediction,which is a research hotspot of large-scale wind power grid integration.This paper adopts an improved multi-position multi-step wind speed prediction method.First,a decomposition and reconstruction strategy of wind speed matrix time-spatial correlation is proposed,and the gray correlation analysis is performed on the wind turbines in the wind farm.Based on this,the improved spatial association algorithm is used to optimize and sort,and obtain spatial information of typical wind turbine and neighboring domains.Information is reconstructed to improve the efficiency of spatial feature extraction.Then,the reconstructed spatio-temporal three-dimensional information is input into the convolutional memory network to reduce the impact of the lack of information on the prediction accuracy,and spatial feature extraction and multi-step ultra-short-term prediction are performed.Finally,the proposed method verify the prediction accuracy and generalization ability by predicting the wind speed and wind power of different wind farms.

关 键 词：多步预测 灰色关联 空间关联优化算法 卷积记忆网络 

分 类 号：TM721[电气工程—电力系统及自动化]