摘要
分别采用线性相关方法、物理方法和支持向量机(SVM)方法对内蒙古不同地形的两个风电场进行了基于邻近测风塔2个月学习数据的10个月平均风速估算研究,给出了较详细的实现过程,并比较了3种估算方法的预测精度。计算结果表明,线性相关方法能快速给出预测结果,但要求两塔之间的空间相关性良好;物理方法对较平坦地形估算结果精度较高,但对于较复杂地形估算结果平均百分比误差高达15%;SVM方法能给出精度较高的估算结果,在两个风电场的估算平均百分比误差均小于5%,对复杂地形适应性强,综合结果好于其他两种方法。
Three methods were investigated to evaluate ten-month wind speed based on the two-month learning data of the neighboring masts, which located in two wind farms with different terrains in Inner Mongolia. The adopted strategies were linear correlation method, physical method and the support vector machine （SVM） method. The performance of three estimation methods were comprehensively compared. The calculated results show that the linear correlation method can yield the estimation results quickly, while the good spatial correlation between the two masts is required ; the physical method performs excellently for the fiat terrain, while its performance is poor for the complex terrain with the mean absolute percentage error （MAPE） of 15% ; the SVM method functions well for the two wind farms with the MAPE less than 5%. Therefore, the SVM method is robust for the complex terrain and exceeds other two methods in a whole.
出处
《太阳能学报》
EI
CAS
CSCD
北大核心
2012年第2期230-235,共6页
Acta Energiae Solaris Sinica
基金
国家自然科学基金重点项目(50837003)
国家重点基础研究发展计划(2009CB219701)
关键词
长期风速估算
线性相关
物理方法
SVM方法
long term wind speed evaluation
linear correlation
physical method
support vector machine method
