Data-driven assessment of the varied effects of the peak time rebate on household electricity usage

Abstract

Recent shifts toward demand-side electricity management have brought increased attention to Peak Time Rebate (PTR) initiatives, which aim to reduce household electricity use during peak hours by offering financial incentives. However, previous studies often overlook the heterogeneity in household responses-that is, differences in how individual households react to these incentives-and the long-term effects of behavioral changes triggered by PTR programs. To address this research gap, this study employs machine learning techniques to analyze hourly electricity consumption for 125 households participating in the People Demand Response (DR) program, a PTR initiative in Korea. First, households are clustered based on their hourly electricity consumption patterns. Machine learning is then used to learn consumption patterns, and a predictive model is applied to evaluate the impact of DR events by estimating the counterfactual condition. The findings indicate varying effects of DR interventions across these clusters. Moreover, learning effects emerged over time within specific clusters, highlighting the need for personalized targeting strategies. This study disputes the universality of PTR impacts and offers guidance for designing more effective and enduring PTR programs by service providers and policymakers.