Two-Stage Clustering of Household Electricity Load Shapes for Improved Temporal Pattern Representation

With the widespread adoption of smart meters in buildings, an unprecedented amount high-resolution energy data is released, which provides opportunities to understand building consumption patterns. Accordingly, research efforts have employed analytics and machine learning methods segment customers based on their load profiles, help utilities providers promote customized/personalized targeting for programs. Existing segmentation techniques use assumptions that could reduce clusters’ quality representing members. Therefore, this paper, we investigated a two-stage clustering method capturing more representative shape temporal patterns peak demands through cluster merging approach. In first stage, shapes are clustered (using classical algorithms) by allowing large number clusters accurately capture variations patterns, centroids extracted accounting limited misalignment within range Demand Response (DR) timeframes. second with similar power magnitude ranges merged using Complexity-Invariant Dynamic Time Warping. We used three datasets consisting ~250 households (~15000 profiles) demonstrate efficacy framework, compared baseline methods, discuss impact management. The proposed merging-based also increased correlation between corresponding members 3–9% different datasets.