区间电力负荷特征提取及预测方法  

A Characteristic Extraction and Forecast Method for Interval Power Loads

作　　者：马立新[1] 李渊[1] 

机构地区：[1]上海理工大学光电信息与计算机工程学院,上海200093

出　　处：《控制工程》2015年第4期645-648,共4页Control Engineering of China

基　　金：国家科技部政府间科技合作项目(2009014);上海市创新基金项目(jwcxsl1302)

摘　　要：电力负荷是典型的时间序列数据,提前对未来一个月每天最大负荷值的预测,是一个有时间间隔的预测过程。对于连续多天负荷预测的研究中,节假日负荷预测是一个难题,需把待预测日分成法定假日和非假日两类分别进行处理。引入区间负荷预测方法,用标幺值量化假日负荷特征和趋势,并通过小波分解来提取非假日负荷变化的特征,分别建立对应的BP神经网络模型进行预测。某市电力负荷数据的预测结果表明,这种预测方法为间隔时间区间的最大负荷预测问题提供了一个有效的解决方案。Power load is typical time-series data. Forecasting the daily peak load in continuous multi-days ahead of time is a time-interval forecast process. In the process of multi-days load forecasting, holiday load forecasting is a thorny problem. To solve this problem the load data should be divided into two parts： national holiday and non-holiday. By introducing interval load forecasting method, quantifying the feature of holiday load by per unit value, and extracting the non-holiday load change through wavelet decomposition, the corresponding BP neural network model is established to forecast separately. The simulation results show that this method provides an effective solution for the interval time-series peak load forecasting.

关 键 词：区间时间序列 特征提取 小波分析 负荷预测 神经网络 

分 类 号：TM715[电气工程—电力系统及自动化]