基于数据挖掘的电力系统中长期负荷预测新方法  

Mid-long term load forecast of power system based on data-mining

作　　者：崔旻[1] 顾洁[1] 

机构地区：[1]上海交通大学电气工程系,上海200240

出　　处：《电力自动化设备》2004年第6期18-21,共4页Electric Power Automation Equipment

摘　　要：中长期电力系统负荷预测受大量不确定因素的影响,研究表明聚类方法能够将各种影响因素综合引入预测模型。所提出的改进聚类算法结合了层次方法中的变色龙(Chameleon)法与基于密度算法的优点,实现了最优聚类,同时还弥补了单纯层次法无法对复杂形状数据聚类和算法不可逆的缺点。算法在进行聚类前以不完备数据分析补全法算法(ROUSTIDA)为数据处理前导,确保了聚类所需历史数据的准确性和完备性。实践证明该算法具有计算速度快、预测精度高、预测误差变化小等优点。尤其在影响因素繁多、历史数据不完整或不准确时,改进算法更能体现出优越性。Mid - long term load forecast of power system is affected by various uncertain factors and research shows that clustering method can synthesize numerous relative factors and put them into forecast model.An improved clustering method is put forward ,which unites the advantages of Chameleon algorithm and density - based algorithms to achieve optimal clustering.It makes up the disadvantages of hierarchy method ,which is irreversible and useless for the complex shapes.With ROUSTIDA algorithm as the pre - process ,exact and intact historical data are ensured.Practical examples prove its fast calculation speed ,high forecast accuracy and small error variety.When there are more in - fluencing factors and the historical data are non - integrated or inexact ,this method shows especially its advantages.

关 键 词：电力系统 中长期负荷预测 数据挖掘 聚类分析 不完备数据分析 

分 类 号：TM715[电气工程—电力系统及自动化]