Because electricity cannot be practically or economically stored in large quantities, the electricity generation and distribution system must match supply and demand on a minute-by-minute basis. Delivery of electricity for residential use has traditionally been done by matching the supply to the demand, with little or no control over the demand. This causes severe distortions in the system operation and economics when the demand hits unusually high peak values. When these peaks are particularly high, some effort to reduce demand is done via roadside signs, television appeals, etc. (the California heat spell of July, 2006, for example).

There is presently considerable effort in the electricity industry and regulation organizations to implement “demand responsive” mechanisms into the system. These would involve signals sent to the residence that would cause reductions in usage through pricing or other means. When such demand response (DR) signals are sent, the system, consisting of large numbers of residences, will respond with dynamic behavior based on the aggregate properties of all of the residences involved. The nature of the dynamic behavior will affect the effectiveness of the DR strategy, for example, peak power usage generated at the end of a DR period, the so-called rebound peak. In this project, we are constructing a model of how a large number of residences would respond to various types of DR signals so various strategies can be compared and effective strategies can be designed.

2009 Update:

Key Achievements and Societal Impact -- Interim Report

This interim report highlights further development on a large scale residential energy consumption simulation used for studying demand response and load management. The simulation models a group of houses, each with independent heating ventilation and air conditioning (HVAC) systems controlled by independent thermostats. The individual house properties are randomly chosen based on house population data, and no two houses are alike. Further, the thermostats are intelligent, in that they have the ability to modify their power consumption in response to communications sent by a central authority.

This work added residents to the simulation in order to facilitate comfort and behavior studies. Now, each house contains one or more residents that interact with the thermostat based on their individual work/home/sleep schedule and comfort preferences. Each resident is modeled using a stochastic state machine with independent preferences chosen based on behavior statistics. The residents virtually come and go, and when they are in the house, they can get uncomfortable enough to change the thermostat set point temperature. Their behavior and interactions are simulated with a degree of randomness, just like real people.

Demand response and other load management techniques have great potential to reduce the cost of energy as well as improve the environmental impact of electricity generation. If used without proper understanding, these technologies also have potential to create catastrophic instabilities in the electricity grid. The energy consumption simulation provides an environment where safe experimentation can be performed before deploying advanced load management technologies. The addition of residents into the simulation improves the accuracy of the model as well as providing more information to help understand the full impact of load management.

Burke, W. J., Peffer, T. and Auslander, D.M., 2009 "UNDER REVIEW: Residential Occupied Neighborhood Simulation -- RON-Sim "ASME Journal of Dynamic Systems and Controls.