智能用电用户行为分析特征优选策略

针对大数据应用背景下用户智能用电行为分类的计算复杂性和特征选择有效性的问题,提出一种基于特征信息量的特征优选策略。首先,以用电特征的互信息量与相关系数作为特征有效性和关联性判据,设计用电特征优选准则。然后,提出一种该准则下的用电行为特征优选策略,通过减少特征间的分类信息冗余实现高维特征的降维,并选取有效独立的特征,从而构建用户用电行为聚类精简特征集。最后,基于特征优选策略实现了一种特征自适应的用户用电行为分析方法,完成优化的用户用电行为分析。通过电网实际用电数据验证了所提策略能提高聚类准确率和减少计算复杂性的有效性。

Feature Selection Strategy for Electricity Consumption Behavior Analysis in Smart Grid

In face of the problem with mass electricity data processing, a strategy of electricity consumption behavior feature selection is put forward to reduce computing complexity and improve validity. Firstly, the validity and the relevance of electricity consumption feature are analyzed by its information entropy and related coefficient. On this basis, an evaluation criterion for electricity consumption feature is designed. Then a feature selection strategy is built to reduce the high dimensions of the feature vectors by reducing the redundancy of feature space. And the condensed feature set is devised using this strategy to select features. According to the feature selection strategy, a method of electricity consumption behavior analysis is put forward to make electricity consumption behavior analysis optimized. Finally, the simulation results of this method achieved by electricity consumption data show that it can effectively improve the accuracy of clustering and reduce computation time.