Discipline: Computer Sciences and Information Management
Subcategory: Computer Science & Information Systems
Ryan Read - New Mexico State University
Co-Author(s): William Kluegel, New Mexico State University, NM; Jeremiah Smith, New Mexico State University, NM; Stepheny Perez, New Mexico State University, NM; Robert Lykins, New Mexico State University, NM; Michael Garcia, New Mexico State University, NM; Ferdinando Fioretto, New Mexico State University, NM; Son Tran, New Mexico State University, NM; William Yeoh, Washington University in St. Louis, MO; Enrico Pontelli, New Mexico State University, NM;
Demand for electricity fluctuates throughout the day, eventually reaching a maximum in the early evening as people return from work. This maximum is known as ‘peak load’ and is the main contributor to many challenges for both consumer and energy company alike. Energy companies must always generate enough electricity to meet demand, or else risk a brownout. It is for this reason that they must run more generators than they predict they will need to meet demand. Most of these extra generators start up quickly but are costly to run, less efficient, and have higher emissions than the generators normally used. Most of the time, this extra energy goes unused. Some energy companies have chosen to implement variable pricing structures to reduce peak load, reducing the number of extra generators they must keep running always. Variable pricing incentivizes users to consume energy during off-peak hours instead of costly peak hours.
In order to help power companies reduce wasted energy as well as reduce energy bills for consumers we propose the Multi-Agent Energy-Efficient Smart Home system (MESH). The idea behind MESH is to reduce peak power consumption by spreading out power intensive tasks throughout the day. Using users’ preferences, a constraint optimization program generates a schedule that minimizes cost and peak power consumption. This schedule is executed using Internet of Things (IoT) enabled devices.
If all households attempt to schedule tasks independently in non-peak hours new peaks are created. As a consequence, rather than flattening the existing peak, it is moved to another time of the day. To address this challenge, MESH implements communication between households. Each house aggregates the intended power consumption of neighboring houses when attempting to reduce peak hour usage. Thus, MESH allows entire neighborhoods to cooperatively plan, save, and help the power companies as well.
Our current prototype consists of a web user interface, a schedule generator, an agent control system, and a device control system. The web interface allows users to input preferences about device usages or desire values of environment. These rules are used by the schedule generator to generate a schedule that minimizes peak hour energy usage and price. The schedule is then passed to the agent control system, where power consumption is shared to neighboring agents. Once a satisfactory schedule is agreed upon, the agent can execute the schedule using the device control system. The device control system can then control a variety of real or simulated devices such as a robotic vacuum, lighting, and a range of sensors.
We also have plans to continue to improve MESH. Our goal for the scheduler is to improve the speed and quality of the solution that it produces. As we improve the scheduler, we hope to be able to handle more complex inputs to ensure user preferences are taken into account. We also hope to increase the number and variety of devices we can control.
Funder Acknowledgement(s): Funding was provided by an NSF grant via the NMSU iCREDITS program.
Faculty Advisor: Tran Cao Son, tson@nmsu.edu
Role: I worked as a software developer to build the MESH system.