海上风电数值模式误差修正技术研究
RESEARCH ON NUMERICAL MODEL ERROR-CORRECTION TECHNIQUES FOR OFFSHORE WIND POWER
张皓 1文仁强 1杨定华 1易侃 1杜梦蛟1
作者信息
- 1. 中国长江三峡集团有限公司科学技术研究院,北京 101100
- 折叠
摘要
针对现有数值模式预报数据精确度有限的问题,结合广东海域多处实际测风数据,分别采用卷积长短期记忆网络、集成学习等多种机器学习框架建立误差修正模型,并对各模型进行适用性验证及分析.在此基础上,提出更适用于风向误差计算的基于三角函数规律的代价函数.结果表明,所建立的AdaBoost及GBDT模型对风速、风向变量的修正均能取得优异效果.
Abstract
In view of the limited accuracy of the existing numerical model prediction data,this paper combined with the actual wind measurement data of several places in the Guangdong sea area,respectively using convolutional long short-term memory network,integrated learning and other machine learning frameworks to establish error correction models,and to verify and analyze the applicability of each model.On this basis,a cost function based on trigonometric function law which is more suitable for wind direction error calculation is proposed.The results show that the AdaBoost and GBDT models can achieve excellent results in the correction of wind speed and wind direction variables.
关键词
海上风电/数值模式/误差修正/机器学习/提升模型Key words
offshore wind power/numerical model/error correction/machine learning/boosting model引用本文复制引用
出版年
2024
NETL
NSTL
万方数据