Consumer profiling for demand response programs in smart grids

In this paper, we propose new models for consumer demand response estimation in a smart energy environment, where consumers have access to real time electricity pricing information and can respond to price signals by changing their energy consumption through a two-way communication system. We introduce a stochastic model that differentiates and characterizes two principal constituents of consumers demand response behavior: a long-term steady behavior and a short-term dynamic response behavior. We further propose a method to estimate conditional probability distributions of future demand given current demand and price information, which gives a complete probabilistic characterization of the short-term dynamic response behavior. This approach extracts much more information on consumer behavior from a given set of data than the traditional approach which estimates statistics such as demand elasticity directly. We demonstrate our methodology with the residential demand response experimental data taken from the Olympic Peninsula project, and discuss in detail the results of the proposed approach.