Decentralized elastic electricity demand scheduling

The adoption of intermittent renewables in the electricity grid will increasingly require loads to track available generation. This paper proposes several algorithms to schedule flexible-power loads that arrive unpredictably to a grid-connected microgrid. These are based on standard schedulers such as Shortest Remaining Processing Time (SRPT), adapted to the fact that appliances have maximum power ratings. A simple decentralized scheduler is applied to ensure that the aggregate load does not exceed the generation, and a gain control mechanism is proposed to stabilize the system. The proposed scheduler has two sources of sub-optimality: determining the values of the control signals, and their overall structures. In order to separate these two effects, we consider a full communication version as a benchmark in order to assess performance against the SRPT-based algorithms and fair sharing. This full communication "decentralized" approach approximates the performance of SRPT-based policies, suggesting that a true decentralized controller may be feasible in the future.