Customer Targeting for Load Flexibility via Resident Behavior Segmentation

Abstract

Caused by the widespread of renewable energy resources, oversupply and overdesign of reserve capacity are serious concerns in electrical power system. Hence, power system needs load flexibility (LF) to mitigate the concerns. Despite utilities’ efforts to aggregate LF for residential loads, aggregated LF operations are typically inefficient due to load uncertainties caused by individual customers’ varying behaviors. Therefore, in this study, a novel customer targeting model for LF participation is proposed based on resident behavior analysis. The proposed scheme comprises a resident behavior segmentation model (RBSM) performing deep-learning-based and probabilistic clustering and a virtual LF operation model (VLFOM) considering human comfort. Simulations were conducted in a residential environment in San Diego, California. The superior performance of RBSM was confirmed through internal clustering evaluation methods. Accordingly, the results of the VLFOM model confirmed the criteria of resident behaviors for adequate LF participation. The experiment results present that the proposed model is economically and environmentally effective to the LF operation for utilities. Finally, implications of LF aggregation based on uncertainty assessment and meaningful suggestions of the settlement design for future LF expansion are presented. IEEE