基于非线性IRWLS-SVM短期电价预测的改进方法  

Short Period Eelectricity Price Forecast Based on Nonlinear Irwls-svm Improved Method

作　　者：马历[1] 马克雄[2] 

机构地区：[1]中国长江三峡集团公司市场营销部,北京100038 [2]三峡大学电气与新能源学院,湖北宜昌443002

出　　处：《三峡大学学报（自然科学版）》2012年第6期39-43,共5页Journal of China Three Gorges University：Natural Sciences

摘　　要：本文提出了一种基于非线性IRWLS-SVM的短期电价预测的模型.首先对不同的损失函数进行仿真,选出具有一定鲁棒性的Huber损失函数来适应模型小的变化.然后比较线性、径向基、多项式3种核函数.仿真结果表明,多项式核函数的预测效果最好.最后提出了一种改进的非线性IRWLS-SVM算法,仿真结果表明改进后的算法提高了局部预测精度.This paper has introduced a short period electricity price forecast model based on nonlinear iterative re-weighted least squares（IRWLS） support vector machine（SVM）. First of all, it has simulated with differ- ent cost functions~ and the result has indicated that Huber cost function has robustness, and it can adapt to small changes of the model. Then, this paper has compared with three kernel functions, i. e. linear kernel function, radial basis kernel function and polynomial kernel function. The simulation results show that the polynomial one has the best regression result. At last, the paper proposes an improved algorithm which can improve the local forecasting precision.

关 键 词：支持向量机 非线性 短期电价预测 

分 类 号：TM715[电气工程—电力系统及自动化]