短期负荷预测的聚类组合和支持向量机方法  

Application of Clustering Combination and Support Vector Machine in Short-term Load Forecasting

作　　者：梁建武[1] 陈祖权[1] 谭海龙[1] 

机构地区：[1]中南大学信息科学与工程学院,长沙410083

出　　处：《电力系统及其自动化学报》2011年第1期34-38,共5页Proceedings of the CSU-EPSA

基　　金：国家自然科学基金资助项目(60173041)

摘　　要：为提高电力系统短期负荷预测的精度,提出了一种基于聚类组合和支持向量机的短期负荷预测方法。该方法用SOM网络训练规格化的特征数据并获得初始聚类中心,将初始聚类中心作为C-均值算法的输入,并用DB指数评价聚类结果以获得最佳聚类数,通过训练可得相似日样本,最后选择合适的参数和核函数构造支持向量机模型来进行逐点负荷预测。预测结果表明,该方法比单一的支持向量机算法具有明显的优势。In order to improve the accuracy of power system short-term load forecasting,a method of short-term load forecasting based on clustering combination and support vector machine is proposed.First,the standardized data are trained through SOM network and the initial clustering center are acquired,then the initial clustering centre is used as the input values of the C-means algorithm,and the best number of the clustering is obtained through DB index,the samples of similar days are acquired through training.Finally,support vector machine using appropriate parmaeters and kernel function are constructed and the load was forecasted point by point.The results showed that the method has a distinct advantage than simple support vector machine algorithms.

关 键 词：短期负荷预测 聚类组合 SOM网络 C-均值 相似日 

分 类 号：TM715[电气工程—电力系统及自动化]