Online Residential Demand Response via Contextual Multi-Armed Bandits

Residential loads have great potential to enhance the efficiency and reliability of electricity systems via demand response (DR) programs. One major challenge in residential DR is how learn handle unknown uncertain customer behaviors. In this letter, we consider problem where load service entity (LSE) aims select an optimal subset customers optimize some performance, such as maximizing expected reduction with a financial budget or minimizing squared deviation from target level. To behaviors influenced by various time-varying environmental factors, formulate contextual multi-armed bandit (MAB) problem, develop online learning selection (OLS) algorithm based on Thompson sampling solve it. This takes information into consideration applicable complicated settings. Numerical simulations are performed demonstrate effectiveness proposed algorithm.