Back to publications

SINERGYM - A machine learning framework for building energy optimization

Campoy-Nieves, Alejandro and Manjavacas, Antonio and Jiménez-Raboso, Javier and Molina-Solana, Miguel and Gómez-Romero, Juan
Building Simulation (under review) (SUBMITTED)

Abstract:

Simulation has become an integral part of Building Energy Optimization (BEO) as it enables testing of different design and control strategies at a low cost. Particularly, Machine Learning (ML) algorithms can leverage large-scale simulations to learn optimal control from vast amounts of data without supervision, as seen in the Deep Reinforcement Learning (DRL) paradigm. Unfortunately, the lack of open and standardized tools has impeded the widespread application of ML to BEO. To address this issue, this paper introduces SINERGYM, a Python-based open-source building simulation, control, and monitoring framework that simplifies BEO experiments’ design and execution. SINERGYM is compatible with EnergyPlus simulation models, providing a consistent interface, predefined benchmarks, support for experiment visualization and replication, and comprehensive documentation in a ready-to-use software library. The paper highlights the main features of SINERGYM compared to other existing frameworks, and its basic use, showing its simplicity and potential for advanced BEO research and development.

Links:

DOI:
PDF:

Bibtex:

@article{Campoy2023,
  author = {Campoy-Nieves, Alejandro and Manjavacas, Antonio and Jiménez-Raboso, Javier and Molina-Solana, Miguel and Gómez-Romero, Juan},
  title = {SINERGYM - A machine learning framework for building energy optimization},
  journal = {Building Simulation},
  year = {SUBMITTED},
  volume = {(under review)},
  doi = {},
  comment = {},
  timestamp = {37}
}