Demand response is an alternative, environmentally friendly source of regulation, in which loads adjust their power consumption to meet power system needs. Two major challenges to demand response are (i) problem scales of up to 10^5 loads, and (ii) uncertainty stemming from a sparse communication infrastructure. In the first part of this talk, we address these issues by modeling demand response as a 'restless bandit' problem, and obtain an index policy solution. The index policy easily scales to millions of loads and nearly optimally balances the classical tradeoff between exploration and exploitation.

In the second part of this talk, we consider the economics of regulation. The recent FERC order #755 has mandated that frequency regulation be made more fair and efficient. We propose letting regulation prices be the dual multipliers or 'costate' of the linear quadratic regulator in the same manner that locational marginal prices are derived from the dual of economic dispatch. We then construct a Vickrey-Clarke-Groves mechanism to induce selfish agents to honestly report their private valuations and costs of regulation.

Host: Scott Backhaus backhaus@lanl.gov