Residential energy demand presents a unique challenge in the overall electric grid. Although residential energy only constitutes 34% of total electricity consumption in the United States , the same demand disproportionately accounts for up to 50% of peak load in certain parts of the country . In hotter climates, this change in load is driven primarily by air-conditioning . The strain on the grid from residential energy necessitates operating additional “peaking plant” facilities for a limited period of time, which are designed for low capital cost rather than high energy efficiency and have a much higher environmental impact than their “base load” counterparts, which are operated continuously .
Until recently, detailed information on residential energy behaviors and patterns was limited due to insufficient data. However with the mass installation of smart meters, meters that measure and record electricity usage in high resolution, there is now a large amount of data available to analyze and predict home energy behaviors. Utilities are interested leveraging smart meter data to isolate and quantify potential changes that residential energy consumers can make in order to stabilize the grid and minimize peak loads. Although it is possible to conduct a detailed model and energy audit for an individual house, the relative benefit of one house is minimal. Instead it is most cost effective to leverage smart meter data to statistically identify energy intensive houses within a broad set of smart meter data and provide general recommendations that result in lowered peak demand
In our work we present an statistical identification method to recognize the most energy intensive houses from a broad data set of smart meter data and recommend operational changes (e.g. thermostat settings or appliance usage) or capital improvements (e.g. installation of thermal insulation or a high-efficiency appliance). An analysis of the data reveals an "energy slope" as a function of ambient temperatures and internal parameters. Household surveys and energy audit data combined with K-means clustering and parametric analysis highlight the most significant components affecting the energy slope. The analysis in turn provides an estimate of the energy reduction when considering multiple mitigation opportunities. The benefit of these suggested changes is many of the operational changes require no additional capital cost and are targeted toward minimal violations of normal preferences. The research also compares the additional benefit of higher-resolution smart meter data. A yearly data set of smart meter data from 100 houses from the Pecan Street project in Austin, TX in 1-minute time intervals provides a case study demonstrating the efficacy of the method.
 U.S. Energy Information Administration. Electric Power Annual Electricity Report. March 23, 2015.
 Wattles, ERCOT Demand Response Overview & Status Report , 2011
 K.X. Perez, et. al. Nonintrusive disaggregation of residential airconditioning loads from sub-hourly smart meter data. Energy and Buildings, 81:316–325, 2014.
 W.J. Cole, et. al. Use of model predictive control to enhance flexibility of thermal storage cooling systems. In Proceedings of American Control Conference, 2012.