基于扩展面板大数据的电力经济特征提取新方法

A Novel Feature Extraction Method in Power Economic Assessment Research Based on Extraction Panel Data

电力作为经济的"示波器",对海量电力用电大数据进行特征提取和智能参数估计是电力经济评估的关键步骤。提出了一种适用于海量电力经济大数据的建模方法和经济相关特征提取方法。首先针对电力经济二元大数据的时空特征构造扩展面板数据模型,并进行平稳性和协整性检验;然后以用电量为因变量,通过构造回归方程确定与其他电力经济特征量的权重因子;最后采用灰色关联聚类进行特征提取,并以权重因子为判据进行聚类中心选择,从而获取最优特征子集。通过对某省实际用电数据的仿真对比分析,验证所提方法能够在保存特征子集物理含义的前提下,极大消除冗余,满足了经济评估的需要,并具有一定的通用性。

As electricity serves as an economic"oscilloscope",feature extraction and intelligent parameter estimation of massive electricity consumption big data are the key steps of power economy evaluation.In this paper,a modeling method and economic related feature extraction method suitable for massive power economic big data are proposed.First,the extended panel data model is constructed according to the spatiotemporal characteristics of the binary big data of electric power economy,and the stationarity and cointegration are tested.Second,taking the power consumption as the dependent variable,the weight factors of other power economic characteristics are determined by constructing regression equation.Finally,the grey relational clustering is used to extract the features,and the weight factor is used as the criterion to select the clustering center,so as to obtain the optimal feature subset.The simulation and comparative analysis of the actual power consumption data in a province shows that the proposed method can greatly eliminate redundancy on the premise of preserving the physical meaning of feature subset,meet the needs of economic evaluation,and have a certain generality.