基于FCM聚类与SVM的电力系统短期负荷预测

Power System Short-term Load Forecasting Based on Fuzzy-C Mean Clustering Algorithm and Support Vector Machines

分析了模糊C均值(FCM)聚类算法,介绍了支持向量机(SVM)回归的基本原理,提出了一种将FCM聚类算法和SVM结合使用的电力系统短期负荷预测方法。该方法考虑到电力负荷变化周期性的特点,通过对学习样本的聚类,选用同类特征数据作为模型的预测输入,然后对各个模型的输入数据进行归一化处理和分类识别,得出最后总的预测输出。此预测方法保证了数据特征的一致性以及算法的全局性,避免了算法陷入局部极小的缺陷。经过仿真实验,证明了该方法的有效性。

The paper analyses the fuzzy C-mean （FCM） clustering algorithm, introduces the principles of support vector machines （SVM） ,and presents a short-term load forecasting method which uses the fuzzy C-mean clustering algorithm and support vector machines conjunctively. In the proposed method ,the periodical feature of power load variation is considered. The homogeneous characteristic data are chosen as the forecasting input of corresponding models by means of the clustering of study samples. Then the input data of each model are normalized and sorted. At last the total forecasting output is get. Thus the consistency of data characteristics and globalization of the algorithm are ensured by this method. Simulation results show the effectiveness of the proposed method.