基于聚类分析的负荷分类和模型泛化能力研究.pdfVIP

基于聚类分析的负荷分类及模型泛化能力研究 郑晓雨 (华北电力大学电气与电子工程学院，北京市，102206) THE RESEARCH ON LOAD CLASSIFICATIN BASED ON CLUSTER ANALYSIS AND MODEL GENERALIZATION ZHENG Xiaoyu (Department of electrical and electronic engineering, North China Electric Power University, Beijing, 102206) ABSTRACT : The time-varied characteristics of the load make 中图分类号：TM714 the load modeling a well-known difficult problem. The key to solve this problem is to find the laws behind the variations of 0. 引言 the load. In this paper, cluster analysis is used to classify the loads. Before that, two eigenvectors used in cluster analysis are 由于电力系统自身的特点决定了对它的研究不 compared. Based on the measured load dynamics in the power 可能通过频繁地在电网中做试验来实现，所以电力 substation, this paper proposes a new load model classification 系统仿真分析就成为电力系统规划设计、调度运行 and generalization method. The load model responses are 控制的重要工具，而其中负荷模型作为电力系统动 utilized as the eigenvector in the load classification algorithm. 态仿真的重要组成部分，对电力系统的动态特性具 Based on the classification, the load model for each class is 有重要影响。它的准确与否直接影响着仿真的结果 built first. Then the important parameters are found to reduce 和以仿真结果为基础而产生的决策方案，进而关系 the number of load classes. Finally a more robust and practical 到决策方案实施所产生的经济效益和社会效益。随 model with good generalization capability is obtained after the 着近年来计算机、负荷记录仪等装置的推广应用， combination of classes, which is also verified by its strong 大量现场实测数据被记录下来，这就为建立准确的 physical interpretation. 负荷模型提供了前提条件。但是由于负荷的组成、 大小与特性时刻都处于变化之中，单纯从负荷建模 KEY WORDS: load modeling, model classification, cluster 方面考虑，为了准确的反映负荷特性就要为每一时 analysis, model generalization 刻准备一个负荷模型，然而这在工程中是不现实的， 工程中的用到的负荷模型即要求有一定的精度又要 摘要: 为