基于神经网络自适应集成的短期负荷预测  

Short Term Load Forecasting Based on Adaptively Integrated Neural Networks

作　　者：苏蕾[1] 郑应文[1] 

机构地区：[1]福州大学自动化研究所,福建福州350002

出　　处：《微计算机信息》2008年第21期172-174,共3页Control & Automation

基　　金：国家自然科学基金项目(协同模式识别方法在智能控制系统中的应用研究;60675058)

摘　　要：针对在电力系统短期负荷预测应用中,单个神经网络存在预测精度较低、预测结果不稳定、泛化能力差的特点,本文提出一种新的基于多神经网络自适应集成的预测模型。通过对某地区的实际负荷数据进行预测分析表明,该方法以很小的运算时间代价、较小的存储空间代价显著地提高了单个网络的预测精度和泛化能力,具有良好的应用价值。Against that in the application of short term load forecasting in power system, single neural networks having problems like low forecasting accuracy, unstable forecasting results and poor generalization ability, the paper proposes a new model based on multi- neural networks that is adaptively integrated. The forecasting experimental results to practical load data of some local area show that this method notably improves the forecasting accuracy and generalization ability of single neural networks, with tiny operation cost and less memory space, so it has good application value.

关 键 词：短期负荷预测 神经网络集成 径向基神经网络 自适应集成 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TP391[自动化与计算机技术—控制科学与工程]