A new mining framework with piecewise symbolic spatial clustering

• Published in: Applied energy Vol. 298; p. 117226
• Main Authors: Fang, Hongliang, Wang, Yan-Wu, Xiao, Jiang-Wen, Cui, Shichang, Qin, Zhaoyu
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 15.09.2021
• Subjects: Clustering; Electricity consumption pattern; Feature selection; Household characteristics; Multifactor analysis; Household characteristics; Feature selection; Multifactor analysis; Clustering; Electricity consumption pattern
• ISSN: 0306-2619; 1872-9118
• DOI: 10.1016/j.apenergy.2021.117226

Shaping residential electricity consumption behaviors can achieve the effect of load shifting to emit less carbon. Electricity consumption pattern is a good way to describe electricity consumption behavior, and is basically determined by household characteristics (HCs). Effective mining of important factors in HCs can help to shape residential electricity consumption behaviors. This paper proposes a new mining framework for discovering the association between HCs and residential electricity consumption patterns. At first, HCs are classified into six categories by the theory of planned behavior based factor classification. Then, typical electricity consumption patterns (TECPs) of each individual resident in different periods divided by seasons and weekday are extracted by a proposed piecewise symbolic spatial clustering with noise algorithm which combines symbolic aggregate approximation and density-based spatial clustering of applications with noise algorithm. Afterward, TECPs are clustered by hierarchical clustering method. To determine the clustering number, six clustering validity indexes are applied to evaluate the clustering results. Finally, a feature-selection-based multifactor analysis method is designed to evaluate 95 factors in HCs impacts on TECPs. Results of the case study using 3117 records including smart meter data and survey data show that most factors of attitudes toward energy category and socio-demographic category are effective factors for predicting TECPs, while the factors in appliance and heating category are non-significant. The proposed mining framework and the findings can provide guidance of shaping grid-friendly electricity consumption behavior. •A mining framework is proposed for factor analysis of electricity consumption.•A symbolic clustering algorithm is proposed for electricity consumption patterns.•A numerical iterative method is designed for time interval division.•Feature selection is used in multifactor analysis for household characteristics.