基于条件互信息的低冗余短期负荷预测特征选择
|
摘要
为避免负荷预测特征集中冗余特征对预测精度的负面影响,降低预测器复杂度,提出一种基于条件互信息(CMI)和高斯过程回归(GPR)的短期负荷预测特征选择方法.首先,为降低建模所用特征量,根据与目标变量具有最大互信息的特征,选取剩余特征中可对目标变量提供最大信息增益的特征,计算CMI值并进行排序;然后,以GPR为预测器,以其预测结果平均绝对百分比误差为决策变量,按照特征CMI值排序顺序,采用序列前向选择方法,确定最优特征子集;最终,以最优特征子集构建GPR预测模型,并与皮尔逊相关系数法(PCC)和互信息(MI)2种特征选择方法分别结合支持向量机和反向传播神经网络开展对比实验.实验结果证明新方法降低了最优特征集合冗余度与预测模型复杂度,且具有更高的预测精度.
In order to avoid the influence of redundancy features which will affect the predict accuracy of load forecasting and reduce the complexity of predictor,a method based on Gaussian Process Regression( GPR)and Conditional Mutual Information( CMI) for feature selection is proposed for short-term load forecasting.Firstly,to decrease the number of features when building a forecasting model,a feature is selected as the one with the highest mutual information( MI) with the target variable.Then,respect to the selected feature,the next feature which adds the largest information to the target variable is selected. The features are ranked in a descending order according to calculating the CMI.Secondly,according to the order of CMI,a sequential forward selection method with Mean Absolute Percentage Error of GPR is utilized for choosing the optimal feature subset.Finally,GPR with the optimal subset is used for building the forecasting model for load prediction,and compared with support vector machine and back propagation neural network combined with Pearson Correlation Coefficient( PCC) and MI respectively.The results show that the new method reduces the redundancy of the optimal feature set and has a higher forecast accuracy with lower complexity of structure.
In order to avoid the influence of redundancy features which will affect the predict accuracy of load forecasting and reduce the complexity of predictor,a method based on Gaussian Process Regression( GPR)and Conditional Mutual Information( CMI) for feature selection is proposed for short-term load forecasting.Firstly,to decrease the number of features when building a forecasting model,a feature is selected as the one with the highest mutual information( MI) with the target variable.Then,respect to the selected feature,the next feature which adds the largest information to the target variable is selected. The features are ranked in a descending order according to calculating the CMI.Secondly,according to the order of CMI,a sequential forward selection method with Mean Absolute Percentage Error of GPR is utilized for choosing the optimal feature subset.Finally,GPR with the optimal subset is used for building the forecasting model for load prediction,and compared with support vector machine and back propagation neural network combined with Pearson Correlation Coefficient( PCC) and MI respectively.The results show that the new method reduces the redundancy of the optimal feature set and has a higher forecast accuracy with lower complexity of structure.
引文
[1]S.Li,P.Wang,L.Goel.A novel wavelet-based ensemble method for short-term load forecasting with hybrid neural networks and feature selection[J].IEEE Transactions on Power Systems,2015,31(3):1-11.
[2]M.Q.Raza,A.Khosravi.A review on artificial intelligence based load demand forecasting techniques for smart grid and buildings[J].Renewable&Sustainable Energy Reviews,2015,50:1352-1372.
[3]康重庆,夏清,刘梅.电力系统负荷预测[M].2版.北京:中国电力出版社,2017:57.
[4]S.Li,P.Wang,L.Goel.Short-term load forecasting by wavelet transform and evolutionary extreme learning machine[J].Electric Power Systems Research,2015,122:96–103.
[5]杨茂,吕天峰,季本明.混沌理论在电力系统负荷预测中应用综述[J].东北电力大学学报,2015,35(3):18-21
[6]J.Wu,J.Wang,H.Lu,et al.Short term load forecasting technique based on the seasonal exponential adjustment method and the regression model[J].Energy Conversion&Management,2013,70(70):1-9.
[7]崔和瑞,彭旭.基于ARIMAX模型的夏季短期电力负荷预测[J].电力系统保护与控制,2015,43(4):108-114.
[8]李啸骢,李春涛,从兰美,等.基于动态权值相似日选取算法的短期负荷预测[J].电力系统保护与控制,2017,45(6):1-8.
[9]G.Cai,W.Wang,J.Lu.A novel hybrid short term load forecasting model considering the error of numerical weather prediction[J].Energies,2016,9(12):994.
[10]M.R.Akbari-Zadeh.A hybrid method based on wavelet,ANN and ARIMA model for short-term load forecasting[J].Journal of Experimental&Theoretical Artificial Intelligence,2014,26(2):167-182.
[11]孟安波,胡函武,刘向东.基于纵横交叉算法优化神经网络的负荷预测模型[J].电力系统保护与控制,2016,44(7):102-106.
[12]J.X.Che,J.Z.Wang,Y.J.Tang.Optimal training subset in a support vector regression electric load forecasting model[J].Applied Soft Computing,2012,12(5):1523-1531.
[13]N.Huang,Z.Hu,G.Cai,et al.Short termelectrical load forecasting using mutual information based feature selection with generalized minimumredundancy and maximum-relevance criteria[J].Entropy,2016,18(9):330.
[14]赵腾,王林童,张焰,等.采用互信息与随机森林算法的用户用电关联因素辨识及用电量预测方法[J].中国电机工程学报,2016,36(3):604-614.
[15]何志昆,刘光斌,赵曦晶,等.高斯过程回归方法综述[J].控制与决策,2013(8):1121-1129.
[16]M.Seeger.Gaussian processes for machine learning[J].International Journal of Neural Systems,2004,14(2):69-106.
[17]甘迪,柯德平,孙元章,等.基于集合经验模式分解和遗传-高斯过程回归的短期风速概率预测[J].电工技术学报,2015,30(11):138-147.
[18]K.G.Sheela,S.N.Deepa.Review on methods to fix number of hidden neurons in neural networks[J].Mathematical Problems in Engineering,2013(6):389-405.
[2]M.Q.Raza,A.Khosravi.A review on artificial intelligence based load demand forecasting techniques for smart grid and buildings[J].Renewable&Sustainable Energy Reviews,2015,50:1352-1372.
[3]康重庆,夏清,刘梅.电力系统负荷预测[M].2版.北京:中国电力出版社,2017:57.
[4]S.Li,P.Wang,L.Goel.Short-term load forecasting by wavelet transform and evolutionary extreme learning machine[J].Electric Power Systems Research,2015,122:96–103.
[5]杨茂,吕天峰,季本明.混沌理论在电力系统负荷预测中应用综述[J].东北电力大学学报,2015,35(3):18-21
[6]J.Wu,J.Wang,H.Lu,et al.Short term load forecasting technique based on the seasonal exponential adjustment method and the regression model[J].Energy Conversion&Management,2013,70(70):1-9.
[7]崔和瑞,彭旭.基于ARIMAX模型的夏季短期电力负荷预测[J].电力系统保护与控制,2015,43(4):108-114.
[8]李啸骢,李春涛,从兰美,等.基于动态权值相似日选取算法的短期负荷预测[J].电力系统保护与控制,2017,45(6):1-8.
[9]G.Cai,W.Wang,J.Lu.A novel hybrid short term load forecasting model considering the error of numerical weather prediction[J].Energies,2016,9(12):994.
[10]M.R.Akbari-Zadeh.A hybrid method based on wavelet,ANN and ARIMA model for short-term load forecasting[J].Journal of Experimental&Theoretical Artificial Intelligence,2014,26(2):167-182.
[11]孟安波,胡函武,刘向东.基于纵横交叉算法优化神经网络的负荷预测模型[J].电力系统保护与控制,2016,44(7):102-106.
[12]J.X.Che,J.Z.Wang,Y.J.Tang.Optimal training subset in a support vector regression electric load forecasting model[J].Applied Soft Computing,2012,12(5):1523-1531.
[13]N.Huang,Z.Hu,G.Cai,et al.Short termelectrical load forecasting using mutual information based feature selection with generalized minimumredundancy and maximum-relevance criteria[J].Entropy,2016,18(9):330.
[14]赵腾,王林童,张焰,等.采用互信息与随机森林算法的用户用电关联因素辨识及用电量预测方法[J].中国电机工程学报,2016,36(3):604-614.
[15]何志昆,刘光斌,赵曦晶,等.高斯过程回归方法综述[J].控制与决策,2013(8):1121-1129.
[16]M.Seeger.Gaussian processes for machine learning[J].International Journal of Neural Systems,2004,14(2):69-106.
[17]甘迪,柯德平,孙元章,等.基于集合经验模式分解和遗传-高斯过程回归的短期风速概率预测[J].电工技术学报,2015,30(11):138-147.
[18]K.G.Sheela,S.N.Deepa.Review on methods to fix number of hidden neurons in neural networks[J].Mathematical Problems in Engineering,2013(6):389-405.