Cross-Domain Feature Extraction-Based Household Characteristics Identification Approach Using Smart Meter Data

Household characteristics play an important role in helping utilities carry out efficient and personalized services. Current methods to obtain such information e.g. survey are usually costly and time-consuming. The widespread installation of smart meters enables the collection of fine-grained residential electricity consumption data, thus making the identification of household characteristics from smart meter data possible. This paper proposes a cross-domain feature extraction-based household characteristics identification approach using smart meter data. Firstly, in addition to conventional time-domain statistical features, several cross-domain features are extracted using discrete wavelet transform. Secondly, random forest algorithm is used to select a subset of important features and remove redundant information contained in the original feature set. Thirdly, support vector machine is used as classifier with the input of the selected features to infer the household characteristics. Finally, case study using the realistic data from Ireland indicates that the proposed approach shows better performance after incorporating the cross-domain features.