基于改进KPCA和GA-BP神经网络的电能质量稳态指标预测
|
摘要
为提高电能质量稳态指标预测精度,以气象因素、有功负荷及历史电能质量数据作为输入变量,提出一种基于改进核主成分分析(KPCA)和遗传算法(GA)优化BP神经网络的电能质量稳态指标预测方法,首先将改进K-means聚类算法与KPCA相结合,通过改进K-means算法将输入变量划分为不同的子类,降低了核矩阵维数;再利用KPCA提取每类输入变量的非线性主成分,简化网络结构;然后分别将每一类中提取的特征作为BP神经网络模型新的输入变量,并结合GA算法优化BP神经网络参数,建立每一类数据的预测模型。算例应用结果表明,该方法的预测精度明显优于传统BP神经网络预测方法和KPCA+BP神经网络预测方法。
In order to improve power quality steady state indices forecasting accuracy,taking meteorological factors,active load and historical power quality data as input variables,this paper proposed a forecasting method of power quality steady state indices based on improved kernel principal component analysis and GA-BP neural network.Firstly,it combine the improved K-means clustering algorithm with the classical KPCA,divided input variables into different subclasses by improved K-means algorithm to reduce the dimension of KPCA kernel matrix.And then the nonlinear principal components of each input variable was extracted by KPCA to simplify the network structure.Secondly,the features extracted from each class were respectively taken as new input variables of BP neural network model,and the GA algorithm was adopted to optimize the parameters of BP neural network for establishing the steady state index of power quality forecasting model of each type of data.The actual example results show that the forecasting accuracy of this method is significantly better than the traditional BP neural network prediction method and KPCA+BP neural network prediction method.
In order to improve power quality steady state indices forecasting accuracy,taking meteorological factors,active load and historical power quality data as input variables,this paper proposed a forecasting method of power quality steady state indices based on improved kernel principal component analysis and GA-BP neural network.Firstly,it combine the improved K-means clustering algorithm with the classical KPCA,divided input variables into different subclasses by improved K-means algorithm to reduce the dimension of KPCA kernel matrix.And then the nonlinear principal components of each input variable was extracted by KPCA to simplify the network structure.Secondly,the features extracted from each class were respectively taken as new input variables of BP neural network model,and the GA algorithm was adopted to optimize the parameters of BP neural network for establishing the steady state index of power quality forecasting model of each type of data.The actual example results show that the forecasting accuracy of this method is significantly better than the traditional BP neural network prediction method and KPCA+BP neural network prediction method.
引文
[1] A Salarvand,B Mirzaeian,M Moallem.Obtaining a Quantitativeindex for Power Quality Evaluation in Competitive Electricity Market[J].IET Generation,Transmission&Distribution,2010,4(7):810-823.
[2]秦浩庭,李群湛,刘燕,等.基于Monte Carlo的电铁电能质量预测方法[J].电力系统保护与控制,2011,39(13):64-70,77.
[3]崔晓飞,马智远,许中,等.基于离散傅里叶分解的电能质量不平衡度指标AR预测方法[J].现代电力,2013,30(6):38-42.
[4]王灿,宋永,宁志毫,等.湖南电网电能质量智能化监测分析系统[J].湖南电力,2017,37(2):56-59.
[5]丁泽俊,刘平,欧阳森,等.电能质量预测与预警机制及其应用[J].电力系统及其自动化学报,2015,27(10):87-92.
[6]李军,常燕芝.基于KPCA-KMPMR的短期风电功率概率预测[J].电力自动化设备,2017,37(2):22-28,36.
[7]张斌,庄池杰,胡军,等.结合降维技术的电力负荷曲线集成聚类算法[J].中国电机工程学报,2015,35(15):3 741-3 749.
[2]秦浩庭,李群湛,刘燕,等.基于Monte Carlo的电铁电能质量预测方法[J].电力系统保护与控制,2011,39(13):64-70,77.
[3]崔晓飞,马智远,许中,等.基于离散傅里叶分解的电能质量不平衡度指标AR预测方法[J].现代电力,2013,30(6):38-42.
[4]王灿,宋永,宁志毫,等.湖南电网电能质量智能化监测分析系统[J].湖南电力,2017,37(2):56-59.
[5]丁泽俊,刘平,欧阳森,等.电能质量预测与预警机制及其应用[J].电力系统及其自动化学报,2015,27(10):87-92.
[6]李军,常燕芝.基于KPCA-KMPMR的短期风电功率概率预测[J].电力自动化设备,2017,37(2):22-28,36.
[7]张斌,庄池杰,胡军,等.结合降维技术的电力负荷曲线集成聚类算法[J].中国电机工程学报,2015,35(15):3 741-3 749.