|
| 摘要: |
| 对传统的2个线性组合预测模型进行了分析,提出了一个新的线性组合预测模型,该模型不要求权系数和为1,而且权重也可以取负值。同时,为提高精度,提出了支持向量机非线性的组合预测模型,该模型实质上是一个非线性回归模型,利用结构风险最小化原则代替传统的经验风险最小化。该模型充分挖掘原始数据和单一预测模型的信息,以单一模型的预测数据作为组合预测样本,选择径向基核函数的支持向量机进行组合预测。同时给出了解决此问题的基于Matlab的支持向量机工具箱的源程序。以美国加州电力日均价为例,与单一预测方法、线性组合预测进行对比,支持向量机非线性的组合预测方法预测比较精确。 |
| 关键词: 组合预测,支持向量机,电价,误差 |
| DOI： |
| 分类号:TP18;F426.61 |
| 基金项目: |
|
| Combined forecasting model of electricity price based on support vector machine |
|
GAO Shang MEI Liang
|
| Abstract: |
| Based on the study of two conventional linear combined forecasting models,a novel linear combined forecasting model is proposed.The sum of its weight coefficients is not necessary to be 1 and its weight coefficient can be minus.Furthermore,a support vector machine nonlinear combined forecasting model is proposed to improve accuracy,which is essentially a nonlinear regression model,applying SRM(Structure Risk Minimization)instead of ERM(Empirical Risk Mini- mization).It fully mines the information of original data and mono forecasting models,takes the results of mono forecasting models as the samples of combined forecasting model and adopts the radial basic kernel function in combined forecasting.The source code based on the support vector machine toolbox of Matlab is given.With California daily electricity price as a study case, compared with mono forecasting models and linear combined forecasting models,the accuracy of support vector machine nonlinear combined forecasting model is higher. |
| Key words: combined forecasting,support vector machine,electricity price,error |
