Low Emission Building Control with Zero-Shot Reinforcement Learning

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/

arXiv commentAccepted at AAAI 2023. Code available via https://enjeeneer.io/projects/pearl/

References in corpus (13)

• Continuous control with deep reinforcement learning
• Practical Bayesian Optimization of Machine Learning Algorithms
• Soft Actor-Critic: Off-Policy Maximum Entropy Deep Reinforcement Learning with a Stochastic Actor
• Zero-Shot Learning Through Cross-Modal Transfer
• Simple and Scalable Predictive Uncertainty Estimation using Deep Ensembles
• A Review of Deep Reinforcement Learning for Smart Building Energy Management
• Gaussian Processes for Big Data
• Reinforcement Learning and Control as Probabilistic Inference: Tutorial and Review
• Deep Reinforcement Learning in a Handful of Trials using Probabilistic Dynamics Models
• Building HVAC Scheduling Using Reinforcement Learning via Neural Network Based Model Approximation
• Model Predictive Path Integral Control using Covariance Variable Importance Sampling
• Temporal Difference Learning for Model Predictive Control
• NeurOpt: Neural network based optimization for building energy management and climate control