支持向量机模型改进及在短期边际电价预测中的应用  

Supporting Improvements of Vector Machine Model and Its Application in a Short-term Marginal Electricity

作　　者：黎特[1] 皮薇[1] 

机构地区：[1]华北电力大学,河北保定071000

出　　处：《国网技术学院学报》2013年第5期1-5,共5页Journal of State Grid Technology College

摘　　要：传统的最小二乘支持向量机模型对训练样本的各个输入点同等看待,各输入向量的贡献度是相同,未对离群点加以考虑。在最小二乘支持向量机模型中引入隶属度的相关理论并建立模糊隶属度函数,对离群点和正常点赋予不同的贡献度,建立了模糊最小二乘支持向量机模型。最后,以美国PJM电力市场的边际电价预测为例,验证了本文模型的预测精度比传统的最小二乘向量机模型高50%左右,且复杂程度基本不变。While using the traditional support vector machine to predict ,because it view each training sample input vector punishment as the same ,so if the outlier points take place in the training sample set,the degree of the influence of the points on predict system will he very large.In order to solve the problem, this paper uses fuzzy math membership theory and support vector machine forecasting techniques combined to create a new forecasting model.At last,the article selects some of the PJM electricity market load marginal price data in the United States to predict.Compared the improved model with the traditional one,the results show that not only the accuracy is about 50% higher, hut also the model complexity is not increased.

关 键 词：模糊最小二乘支持向量机 模糊集合 隶属度 边际电价 

分 类 号：TM715[电气工程—电力系统及自动化]