SOM神经网络和C-均值法在负荷分类中的应用  

Application of SOM Neural Network and C Means Method in Load Classification

作　　者：王文生[1] 王进[1] 王科文[2] 

机构地区：[1]长沙理工大学电气与信息工程学院,长沙410114 [2]湛江中心人民医院,湛江524037

出　　处：《电力系统及其自动化学报》2011年第4期36-39,共4页Proceedings of the CSU-EPSA

基　　金：湖南省高校创新平台开放基金资助项目(10JJ1010)

摘　　要：负荷时变性和分散性已经成为制约负荷模型推广应用的主要因素,而负荷特性分类则是解决这个问题的有效途径。文中提出基于SOM神经网络的C-均值聚类算法的新的负荷分类方法:以负荷模型参数作为负荷动态特性分类特征向量,应用SOM神经网络对初始训练样本进行分类,将获得的聚类数目和各类中心点作为C-均值算法的初始输入进一步聚类。最后通过动模实验的分类结果表明该方法可自动获取分类数,应用于负荷特性分类研究中具有较强的实用性和有效性。The time variable and dispersive characteristics of load limit the application of the load model, and the load characteristics classification is an effective approach to solve this problem. In this paper, a new classification method based on SOM neural network and C means algorithm was proposed. Load model parameters are chosen as the characteristic vectors in the cluster analysis. The initial training samples are classified using SOM neural network, and the obtained clustering number and clustering center are given as initial value of C means method for further clustering. Simulation results show that the proposed method can automatically obtain the best classification number, and it is practical and effective for load characteristics classification.

关 键 词：电力系统 负荷建模 负荷特性分类 自组织特征映射 SOM神经网络 C-均值法 

分 类 号：TM714[电气工程—电力系统及自动化]