EPIC Spotlight: Students team up to create artificial intelligence for energy efficiency

A team of students making up the first cohort with UMaine’s Experiential Learning Innovation Central (EPIC) used physical computing to propose a smart climate control system for buildings. The project was guided by professor of electrical and computer engineering Ali Abedi, Drew Hooke, and Sean Taylor of the Innovative Media Research and Commercialization Center (IMRC). The student team included Noah Lambert, an incoming first-year in computer engineering; Sam Foglio, a junior in civil engineering; and Delaney Smart, an incoming first-year in Biomedical Engineering. “The main goal was to make something that was a cheaper, more effective solution for the current time until people have the funds to get these big thermostat systems,” says Foglio.

According to the students’ research, most current thermostat systems are programmed with a “set it and forget it” interaction, which may not be the most efficient use of energy resources. The system the students proposed would instead change dynamically according to fluctuations in outside temperature and humidity, and would also determine whether or not people are occupying a space in the building and heat or cool the space accordingly. If there are no people present in a building, the system would only maintain the building’s essential operating equipment.

“It [the climate control system] would take in all those inputs and then it would create a plan to heat the building to use the least amount of electricity and get the most out of it,” says Lambert.

The WiSe-Net Laboratory provided wireless programmable radios, microprocessors, and sensors for the students to experiment with prototyping their ideas. The students proposed AI Climate Control in three parts: Machine Learning, Calculation, and Execution. The system would adapt over time by reading inputs, recognizing patterns, and noticing any changes in expected cycles, then it would use this information to predict and calculate the most efficient schedule to operate a building’s climate control.

Creating a “predicted” temperature schedule is a helpful feature, combined with motion sensors to detect if people are in a room. If every day someone turns the heat up in the morning once they arrive at the building, and turns it down when they leave at night, the AI within the students’ proposed system will begin to follow these patterns and plan for them accordingly. So if someone arrives in the building at 8 am, the system would automatically turn up the heat sooner than it had planned in order to accommodate that person and their habits.

The goal of the proposed product is to keep costs low compared to other smart thermostats. The design can be retrofitted into current heating systems, making it cost-effective and scalable to buildings of any size. The students hope to promote more energy-efficient heating systems in the near future.

Contact: research@umaine.edu

Written by Clarisa Diaz