摘要
电力负荷预测作为电力生产部门的重要工作之一,是根据电力负荷历史数据和其他各类相关影响因素进行预测的,因此,其预测精确度很大程度上取决于历史数据的准确程度。数据挖掘为分析各种海量的、复杂的、含有噪声的数据提供了新的方法。针对电力系统的基本特征,提出一种基于最优区间分割和单调递减阈值函数的聚类方法,然后应用Kohonen网提取相关负荷的特征曲线,并将其用于不良数据的校正。通过对电力负荷的仿真分析验证了该算法的有效性。
Load forecasting is one of the critical jobs in electric power production department. It is based on historical load data and other various correlative-influencing factors. Thus the forecasting accuracy greatly depends on precision of the historical data. Data mining provides a new method for analyzing huge amount of complicated data containing noise. Aiming at the essential characteristics of power system, the clustering method based on optimal regional break up and monotonous decreasing threshold function is proposed. Then correlative load curve is extracted, and the bad data are corrected by adopting Kohonen network. Through simulation analysis, the effectiveness of the method is verified.
出处
《自动化仪表》
CAS
北大核心
2009年第9期14-17,共4页
Process Automation Instrumentation
作者简介
刘齐更，男，1980年生，现为新疆大学电力系统及其自动化专业在读硕士研究生：主要研究方向为电力系统负荷预测.
