摘要
针对移动最小二乘法(MLS)存在的过拟合问题,以及传统正则化方法普遍采用单一的正则项系数,没有充分考虑不同阶次MLS基函数对过拟合产生不同程度的影响,文章提出了一种改进型正则化MLS方法,对不同阶次项采用了不同的正则项系数。将该方法应用于风电场风速风向概率分布的三维曲面拟合。该方法有效克服了传统MLS方法的过拟合问题;拟合后,所显示的风概率分布规律性得到增强;对3个不同区域风分布的拟合结果表明该方法具有较强的通用性。
Aiming at the overfitting problem of the Moving Least Square(MLS)method,and the traditional regularization method generally uses a single regular term coefficient,which does not fully consider that the different orders of MLS basis function brings different degrees of over fitting.Thereforce,an improved regularization MLS method is prposed in this paper,which uses different regularization coeficients for different order terms of the basis.The method is applied to the three dimensional surface fitting of the probability distribution of wind speed and direction in wind farms.The results show that the method effectively overcomes the over fitting problem of the tradition MLS method;By fitting,the performance of the displayed wind probability distribution is enhanced;The fitting results of wind distribution in three different regions show that the method has strong generality.
作者
张丽霞
赵骞
李国良
Zhang Lixia;Zhao Qian;Li Guoliang(Liaoning Provincial Big Data Management Center(Liaoning Provincial Information Center),Shenyang 110002,China;School of Science,Shenyang University of Technology,Shenyang 110870,China)
出处
《可再生能源》
CAS
CSCD
北大核心
2023年第1期67-73,共7页
Renewable Energy Resources
基金
世界银行市场伙伴准备基金项目(P145586)
辽宁省教育厅高等学校基本科研项目(LJKZ0159)
辽宁省教育厅科学研究经费项目(WQN202022)。
关键词
移动最小二乘法
曲面拟合
联合概率密度
改进型
正则化
数值计算
moving least squares
surface fitting
join probability density
improved
regularization
numeral calculation
