摘要
针对基于反向传播神经网络(Back-Propagation Neural Network,BPNN)的中长期电力负荷预测算法中,预测模型的精度和泛化能力易受输入样本变量影响这一问题,利用主元分析(Principal Component Analysis,PCA)方法能消除变量间相关性的特点,对BPNN的输入空间进行重构,消除重叠信息,提取主导因素,优化了网络结构,提高了预测精度。通过实例验证了该方法的有效性。此方法可以使用电计划部门实时、准确的预测电力负荷,以此最优的配比发电机组,也可减少由于预测不准确带来的电力系统各种故障的发生。
To the problem that the precision and generalization performance of forecast model is affected easily by input variable,the method reconstructing the original input space of back-propagation neural network by principal component analysis is presented to eliminate the relevance of value is researched.The proposed method can reduce the duplicated information and extract the leading factors.Its can also optimize its network structure as well as enhance the network forecast precision.The effectiveness of the proposed algorithm is verified by the practical data.The method can be used to real-time forecast power load accurately by the electricity plan division so that power sets can be coordinated reasonably and various faults of power systems resulted by inaccurate forecasting can be reduced.
出处
《控制工程》
CSCD
北大核心
2010年第6期800-802,共3页
Control Engineering of China
基金
国家自然科学基金资助项目(70171021)
国家创新研究科学基金资助项目(60821063)
国家教育部博士点基金资助项目(20080145000)
关键词
主元分析
BP神经网络
负荷预测
电力系统
principal component analysis
back-propagation neural network(BPNN)
load forecasting
power system