Aggregating residential demands with a multi-armed bandit approach

Due to the uncertainty from end users, conventional methods can hardly guarantee the performance while aggregating residential demands. In recent years, the development on the technologies of Internet of things and machine learning create the opportunities to design new algorithms which can tackle users' uncertain behavior. Thus, this paper describes a modern framework of residential demand aggregation, implements a multi-armed bandit approach called CUCB-Avg which was proposed by previous literature, and demonstrates the effectiveness of CUCB-Avg by numerical case studies.