Privacy-Preserving Hybrid Cloud Framework for Real-Time TCL-Based Demand Response

Abstract

Widespread advanced metering infrastructure and wide-area monitoring systems generate a significant amount of electricity load consumption data, which can facilitate eliciting end users temperature flexibility for demand response programs. However, the direct delivery of users load profiles is a threat to users privacy. So this paper proposes a privacy-preserving hybrid cloud framework for TCL-based demand response programs, composed of user private clouds and aggregation cloud. User clouds store users load profiles and elicit temperature flexibility by the proposed stable temperature-related regression model. In the aggregation cloud, this paper proposes the slope-priority flexibility aggregation method for the mean-variance analysis of aggregate flexibility and the XGBoost-accelerated disaggregation model for real-time selecting users based on users fitting coefficients. Hybrid cloud achieves privacy-preserving by separating flexibility eliciting models and aggregation/disaggregation methods into user private clouds and aggregation cloud. Numerical experiments verify that: 1) in user clouds, the stable regression model achieves less predict errors; 2) in aggregation cloud, the slope-priority method can achieve higher aggregate flexibility, and XGBoost-accelerated disaggregating reduces the solving time by nearly three orders of magnitude.