基于克隆选择算法的支持向量回归实现年用电量预测  

Using support vector regression based on clonal selection algorithm in annual electric consumption forecasting

作　　者：孙成发[1] 高辉[1] 

机构地区：[1]阿城继电器股份有限公司哈尔滨瑞雷电气科技发展有限责任公司,黑龙江哈尔滨150090

出　　处：《电力系统保护与控制》2008年第16期11-15,共5页Power System Protection and Control

摘　　要：建立在统计学习理论(SLT)和结构风险最小化(SRM)准则基础上的支持向量回归(SVR)是处理小样本数据回归问题的有利工具,SVR的参数取值直接影响其学习性能和泛化能力。文中将SVR参数选取看作参数的组合优化问题,采用克隆选择算法(CSA)求解该组合优化问题进而选取SVR参数,并应用基于CSA的SVR求解年电力需求预测问题,同时与BP网络预测方法进行了对比。预测结果表明提出的预测方法不仅易于实现,而且精度较高,且性能明显优于BP网络方法。Support vector regression （SVR） is based on statistical learning theory （SLT） and structural risk minimization （SRM） principle, is a powerful tool of solving a small-sample regression problem, and selecting approporiate parameters are very crucial to learning accuracy and generalization ability of SVR. In this paper, the seletion problem of SVR parameters is considered as a combinatorial optimization problem, clonal selection algorithm （CSA） is employed to solve this optimization problem, further, to select approporiate parameters of SVR, and SVR based on CSA is used to annual electric consumption forecasting problem, at same time, comparing forecasting methods based on BP networks is studied. The forecasting example shows that the performance of the proposed method is superior to that of BP networks method in terms of easier realization, higher forecasting accuracy.

关 键 词：年用电量 支持向量回归 克隆选择算法 回归 预测 

分 类 号：TM715[电气工程—电力系统及自动化]