Traditionally, utility companies have employed demand response for large loads or deployed centralized energy storage to alleviate the effects of peak demand on the grid. The advent of Internet of Things (IoT) and the proliferation of networked energy devices have opened up new opportunities for coordinated control of smaller residential loads at large scales to achieve similar benefits. In this paper, we present VPeak, an approach that uses residential loads volunteered by their owners for coordinated control by a utility for grid optimizations. Since the use of volunteer resources comes with hard limits on how frequently they can be used by a remote utility, we present machine learning techniques for carefully selecting which days to operate these loads based on expected peak demand. VPeak uses a distributed and heterogeneous pool of volunteer loads to implement flexible peak shaving that can either selectively target hotspots within the distribution network or perform grid-wide peak shaving. Our results show that VPeak is able to shave up to 26% of the total demand when selectively shaving peaks at local hotspots and up to 46.7% of the demand for grid-wide peak shaving.
Journal or Book Title
Proceedings of the 8th ACM International Conference on Systems for Energy-Efficient Buildings, Cities, and Transportation