Georgia Institute of Technology

Team Members:

Sarah Canastra – Undergraduate Majoring in Industrial Systems Engineering

Hunter Hancock – Undergraduate Majoring in Computer Science (Intelligence and People)

Lucas Kiefer – Undergraduate Majoring in Computer Science (Theory and Intelligence)

School: Georgia Institute of Technology

Challenge: Grid-Interactive Efficient Buildings (GEB)

Problem Definition: Develop conceptual designs that support BTO’s overall GEB strategy in the areas of 1) intelligent algorithms that optimize the operation of building’s active and passive systems to maximize energy efficiency, and 2) whole-building-level interoperable and low cost automation systems that enable communication with building equipment and appliance to optimize operation to provide grid services.

Project Title: Load Shifting with Smart Water Heaters

Solution: With COVID-19 causing a large portion of the population to work from home, energy usage patterns across the country have changed drastically. Increased residential usage during the day has driven larger peaks than ever before, highlighting a clear need to shift the load. Water heaters are the most widely available energy storage mechanism in most homes, presenting a significant opportunity for load shifting when intelligently managed. With 19% of residential energy consumption coming from water heaters that often activate at overlapping times, the entire grid suffers from this lack of management. Among this group, households of 1-2 members make up a majority of the oldest and largest water heaters, contributing disproportionately to energy waste and providing a clear target for our solution. We propose a randomized controlled trial that combines predictive algorithms for water use with direct user input, using deep reinforcement learning to understand the best time to heat the water in each home. Simultaneously, we propose a solution for tracking user preferences through a specially designed mobile application, allowing the needs of the grid and the individual to be intelligently balanced. With the growing prevalence of Time of Use energy plans, this solution would cost less for the end user and reduce energy providers’ need to fall back to less efficient sources of energy during peak hours. Using our algorithm, we were able to reduce peak energy usage due to water heaters by 46% on average, while also keeping the user’s energy costs the same.