摘要
针对电压偏差预测难度大的问题,文中提出一种新的电压偏差预测方法。该方法包括主成分分析法(principal component analysis,PCA)降维、亲和力传播(affinity propagation,AP)聚类、反向传播(back propagation,BP)神经网络预测3步。通过PCA对数据进行降维,获得数据主成分;为了弥补传统聚类方法的不足,提高聚类效果,文中引入AP聚类提取与待预测点同类的历史数据;最后选择BP神经网络建立电压偏差预测模型。将文中方法应用于实际电压偏差数据,结果表明该方法预测结果平均相对误差为3.06%,优于传统BP神经网络预测模型以及BP神经网络结合PCA降维的预测模型。
This paper presents an accurate model to forecast voltage deviation with improved BP neural network,which concerns with the meteorological factors. The proposed method is a combination of PCA dimension reduction,AP clustering and BP neural network. The proposed method is successfully applied to actual data and the practical application results proved that the mean absolute percentage error（ MAPE） of the proposed method is 3.06%,which is obviously better than that of other methods.
作者
王知芳
杨秀
潘爱强
WANG Zhifang;YANG Xiu;PAN Aiqiang(College of Electric Engineering,Shanghai University of Electric Power,Shanghai 200090,China;State Grid Shanghai Electric Power Research Institute,Shanghai 200437,China)
出处
《电力工程技术》
2018年第5期26-31,共6页
Electric Power Engineering Technology
基金
上海市科委地方能力建设计划项目(16020500-900)
关键词
电压偏差
主成分分析
聚类算法
神经网络
voltage deviation
principal component analysis（PCA）
clustering algorithm
BP neural network
