Demand Response in Smart Grids: A Randomized Auction Approach

The smart grid is a modern power grid that achieves high efficiency and robustness through sophisticated information and communications technology. Demand response has great potential in helping balance demand and supply in a smart grid, cutting generation cost and carbon footprint, and improving system stability. Auctions represent a natural and efficient approach for carrying out demand response between the power grid and large electricity users, microgrids, and electricity storage devices. This work explores the modeling and design space of demand response auctions, targeting expressive power, truthful information revelation, computational efficiency, and economic efficiency. We present a randomized auction that explores the underlying problem structure of demand response, and prove that it is truthful, runs in polynomial time, and achieves (1 + ϵ)-optimal social cost for an arbitrarily small constant ϵ. The key technique lies in the marriage of smoothed analysis and randomized reduction, which makes its debut in this work among literature on mechanism design, and can be applied to problems where social welfare optimization is NP-hard but admits a smoothed polynomial-time algorithm.