基于小波变换的支持向量机短期负荷预测  

Short-term load forecasting method based on support vector machine combined with wavelet transform

作　　者：高荣[1] 刘晓华[1] 

机构地区：[1]烟台师范学院数学与信息学院,山东烟台264025

出　　处：《山东大学学报（工学版）》2005年第3期115-118,共4页Journal of Shandong University（Engineering Science）

基　　金：山东省教育厅科技攻关项目(03C03);山东省自然科学基金项目(L2003G01)

摘　　要：提出了一种基于小波分解和支持向量机的短期负荷预测方法.首先利用小波变换把负荷序列分解成不同频段的子序列,对高频序列利用软阀值消噪法去除负荷噪声;对降噪后的负荷序列利用不同的小波进行分解.然后用相匹配的支持向量机模型预测各子序列.仿真结果表明db4小波的预测精度最高,平均绝对预测误差为1.6692%.所得结果同直接用支持向量机预测结果进行比较表明,该方法是有效的.A method based on wavelet decomposing and support vector machine was proposed. Load series was decomposed into different frequency sub-series. The series with high frequency was denoised using soft threshold. Denoised series was decomposed using different wavelet, each sub-series was modeled by using matching support vector machines. Results showed that db4 wavelet had high forecasting accuracy, its absolute percent forecasting error was 1.6692.The result obtained by using support vector machine directly and the above result were compared, simulation showed the effectiveness of the proposed method.

关 键 词：小波变换 支持向量机 核函数 负荷预测 

分 类 号：TM714[电气工程—电力系统及自动化]