欢迎访问《太阳能学报》官方网站,今天是 分享到:
ISSN 0254-0096 CN 11-2082/K

太阳能学报 ›› 2022, Vol. 43 ›› Issue (12): 392-398.DOI: 10.19912/j.0254-0096.tynxb.2021-0597

• • 上一篇    下一篇

基于奇异值分解与卡尔曼滤波修正多位置NWP的短期风电功率预测

王丽婕1, 刘田梦2, 王勃3, 郝颖1, 王铮3, 张元鹏4   

  1. 1.北京信息科技大学自动化学院,北京 100192;
    2.国网北京昌平供电公司,北京 102200;
    3.中国电力科学研究院有限公司,北京 100192;
    4.国网山东省电力公司,济南 250001
  • 收稿日期:2021-05-31 出版日期:2022-12-28 发布日期:2023-06-28
  • 通讯作者: 王丽婕(1983—),女,博士、副教授,主要从事新能源功率预测方面的研究。wanglijie_0203@126.com
  • 基金资助:
    国家自然科学基金(51607009); 北京市属高校高水平教师队伍建设支持计划青年拔尖人才培育计划(CIT&TCD201804053); 北京信息科技大学促进高校内涵发展科研水平提高项目(2020KYNH211); 北京信息科技大学科研基金(2021XJJ17)

SHORT-TERM WIND POWER PREDICTION BASED ON SVD AND KALMAN FILTER CORRECTION OF MULTI-POSITION NWP

Wang Lijie1, Liu Tianmeng2, Wang Bo3, Hao Ying1, Wang Zheng3, Zhang Yuanpeng4   

  1. 1. School of Automation, Beijing Information Science & Technology University, Beijing 100192, China;
    2. State Grid Beijing Changping Power Supply Company, Beijing 102200, China;
    3. China Electric Power Research Institute, Beijing 100192, China;
    4. State Grid Shandong Electric Power Company, Jinan 250001, China
  • Received:2021-05-31 Online:2022-12-28 Published:2023-06-28

摘要: 考虑到数值天气预报网格点位置和系统误差对短期风电功率预测精度的影响,提出一种基于奇异值分解与卡尔曼滤波修正多位置数值天气预报的短期风电功率预测模型。首先通过奇异值分解对多位置数值天气预报数据进行特征提取与降维处理;然后使用卡尔曼滤波方法修正数值天气预报风速数据,降低数值天气预报的系统误差;最后基于极端随机森林算法,利用修正的数值天气预报数据搭建短期风电功率预测模型。通过对某风电场进行仿真,并与单位置、未降维、未修正模型比较,结果表明降维修正模型的预测效果最好,平均误差和均方根误差分别为7.94%和9.96%。

关键词: 风电功率预测, 数值天气预报, 奇异值分解, 卡尔曼滤波, 极端随机森林

Abstract: In consideration of the influences of positional and systematic errors in numerical weather prediction (NWP) grid point on the short-term wind power prediction accuracy, this paper puts forward a short-term wind power prediction model for the correction of multi-position NWP based on the singular value decomposition (SVD) and Kalman filtering, which firstly conducts the feature extraction as well as dimension-reduction process on the multi-position NWP, uses Kalman filtering method to correct the data of wind speed in NWP and to reduce the systematic error of NWP, and finally uses the corrected NWP data to build the short-term wind power prediction model based on the extreme random forest algorithm. Through the simulation of one wind farm as well as the comparison with single-position, non-dimension-reduction and uncorrected models, the results indicate that the dimension-reduction and corrected models have the best prediction effects, and the average error and root-mean-square error (RMSE) are 7.94% and 9.96%, respectively.

Key words: wind power prediction, numerical weather prediction, singular value decomposition, Kalman filter, extreme random forest

中图分类号: