(308f) Flexible Residential Energy: Making the Case for Home Energy Management

Peak power loads in the United States are a costly problem. Of the total electricity generating capacity in the United States, 20% is dedicated to meet peak loads. In other words, 20% of U.S. energy generators are used only operated 5% of the time [1]. It is expensive to install additional generating capacity to guard against large fluctuations in demand. Strategies to mitigate volatility in energy consumption have the potential to reduce the need for surplus capacity. Additionally, even if strategies marginally increase electricity loads, so long as they level the overall load they can increase grid stability and allow for the use of more efficient base power plants. 

Residential consumers are a significant contributor to the peak power problem because of their highly variable nature and dependence on weather. In this work, we investigate the potential for residential consumers to lower peak demand. We also show that home energy management systems with optimal controllers can reduce peak energy consumption for a community of homes while still respecting the preferences of individual users.

In our methodology we will consider the control of the largest loads contributing to peak power consumption: air-conditioning (A/C) and domestic appliances. A/C energy use correlates strongly with weather changes and is the largest load in residential homes. Appliances, such as dishwashers, individually are much smaller than A/C loads. However, the compounding effect of appliances operating simultaneously has the potential to create higher or even new power demand peaks.

First, we present a technique to derive simplified linear models of A/C energy as a function of outdoor dry-bulb temperature and indoor thermostat set points. The development of these models is based on smart meter and smart thermostat data. A non-intrusive load monitoring technique is used to disaggregate A/C electricity consumption from whole-house electricity data reported by the smart meters. A centralized model predictive control (MPC) scheme minimizes peak energy use at the neighborhood level by altering the thermostat set-points in individual homes, while respecting thermal preferences.

After the thermostat settings are planned for the day, we use a novel approach to schedule the operation of appliances, with the goal of minimizing peak load for the community. Dishwashers and washing machines/dryers were identified as potential time-shiftable appliances and are modeled from the probability distributions of start times and required completion times, as determined from consumer survey data. The scheduling problem is formulated as a mixed-integer linear program (MILP) aimed at minimizing peak load under constraints that reflect the start times and allowed delays of individual appliances in each house.

Using sample data collected from forty residential homes located in Austin, TX, we show that the proposed scheduling/control approach can minimize the peak load for the neighborhood by leveraging the physical differences and individual preferences between houses. On average, the centralized controller is able to reduce the daily peak load for the group of houses by 25% (15 kW) when compared with the original settings of the thermostat and appliance start times. In addition, the centralized controller does not significantly increase the total energy consumption and leads to a much flatter consumption profile.

1. H. Farhangi, “The path of the smart grid,” IEEE Power Energy Mag., vol. 8, no. 1, pp. 18–28, Jan. 2010.