Using Meta Reinforcement Learning to Bridge the Gap between Simulation and Experiment in Energy Demand Response
This work addresses data efficiency for energy management in buildings, but it is incremental as it applies existing meta-learning methods to a specific domain.
The authors tackled the problem of costly data collection for reinforcement learning in energy demand response by using meta-learning to warm start experiments with simulated tasks, resulting in improved sample efficiency and better learning with increased complexity.
Our team is proposing to run a full-scale energy demand response experiment in an office building. Although this is an exciting endeavor which will provide value to the community, collecting training data for the reinforcement learning agent is costly and will be limited. In this work, we apply a meta-learning architecture to warm start the experiment with simulated tasks, to increase sample efficiency. We present results that demonstrate a similar a step up in complexity still corresponds with better learning.