局域支持向量回归与误差区间估计的概率预报方法及其应用研究  

Probabilistic Prediction Approach Based on Local Support Vector Regression and Interval Estimation of Its Error

作　　者：张永明[1] 邓盛川[1] 齐维贵[1] 

机构地区：[1]哈尔滨工业大学电气工程及自动化学院,黑龙江哈尔滨150001

出　　处：《电子学报》2010年第1期190-194,共5页Acta Electronica Sinica

基　　金：国家"十一五"重点科技攻关项目(No.2006BAJ01A04);哈尔滨市科技创新人才研究专项资金(No.2006RFXXG010)

摘　　要：针对工程应用对预报技术要求的不断提高以及现有预报方法存在的问题,本文提出了一种局域支持向量回归(Local Support Vector Regression,LSVR)和误差区间估计相结合的概率预报方法,该方法利用局域支持向量回归降低噪声对点预报的干扰,提高预报的可靠性和准确性,利用非参数核估计获取误差区间,避免误差概率分布特性假设,再将点预报和误差区间结合获得预报置信区间,并进一步给出了联合预报置信区间.最后,给出了电网负荷预报和供热负荷预报算例,验证了所提出方法的有效性和实用性.With the increasing requirements of the predicting technologies for engineering,in this paper,a probabilistic prediction approach based on Local Support Vector Regression（LSVR） and interval estimation of its error is proposed to cope with the shortage of existing prediction methods.By means of the proposed approach,LSVR model is used in point prediction to suppress noise interference,while the prediction reliability and accuracy could be improved,and the errors intervals,which avoid the distributional assumptions,could be gained by using nonparametric kernel estimation to the forecast errors.Then combining the point prediction results and errors intervals,the forecast confidence intervals are obtained.Furthermore,joint forecast confidence intervals are also proposed.Finally,the proposed model is performed through simulations by applying it to the data from a real power system and a district heat supply system.

关 键 词：概率预报 局域支持向量回归 非参数核估计 置信区间 负荷预报 

分 类 号：TM714[电气工程—电力系统及自动化]