Low Emission Building Control with Zero-Shot Reinforcement Learning

TL;DR

This paper introduces PEARL, a zero-shot reinforcement learning approach that quickly learns emission-reducing building control policies without prior building-specific data, outperforming traditional rule-based and RL methods.

Contribution

The paper presents PEARL, a novel zero-shot RL method combining system identification and model-based RL to efficiently reduce building emissions without prior data.

Findings

PEARL reduces building emissions by up to 31%.

PEARL outperforms existing rule-based controllers and RL baselines.

A short exploration phase suffices to learn effective control policies.

Abstract

Heating and cooling systems in buildings account for 31% of global energy use, much of which are regulated by Rule Based Controllers (RBCs) that neither maximise energy efficiency nor minimise emissions by interacting optimally with the grid. Control via Reinforcement Learning (RL) has been shown to significantly improve building energy efficiency, but existing solutions require access to building-specific simulators or data that cannot be expected for every building in the world. In response, we show it is possible to obtain emission-reducing policies without such knowledge a priori--a paradigm we call zero-shot building control. We combine ideas from system identification and model-based RL to create PEARL (Probabilistic Emission-Abating Reinforcement Learning) and show that a short period of active exploration is all that is required to build a performant model. In experiments across three varied building energy simulations, we show PEARL outperforms an existing RBC once, and popular RL baselines in all cases, reducing building emissions by as much as 31% whilst maintaining thermal comfort. Our source code is available online via https://enjeeneer.io/projects/pearl/