Crowd-Sourced Storage-Assisted Demand Response in Microgrids

This paper studies the problem of utilizing heterogeneous energy storage systems, including electric vehicles and residential batteries, to perform demand-response in microgrids. The objective is to minimize the operational cost while fulfilling the demand-response requirement. The design space is to select and schedule a subset of available storage devices that are heterogeneous in operating cost, capacity, and availability in time. Designing a performance-optimized solution, however, is challenging due to the combinatorial nature of the problem with mixed packing and covering constraints, and the essential need for online solution design in practical scenarios where both demand-response requirement and the profile of user-owned storage systems arrive online. We tackle these challenges and design several online algorithms by leveraging a recent theoretical computer science technique which uses a problem-specific exponential potential function to solve online mixed packing and covering problems. We show that the fractional version of the algorithm achieves a logarithmic bi-criteria competitive ratio. Empirical trace-driven experiments demonstrate that our algorithms perform much better than the theoretical bounds and achieve close-to-optimal performance.