Self-adaptive building energy control in smart grids

NeurIPS 2019 Workshop Tackling Climate Change with Machine Learning


Buildings are the largest energy-demanding sector in the world, representing over one third of the total worldwide consumption and a similarly important source of CO2 emissions.

We envision a future energy system in which building control will be performed by autonomous self-adaptive agents that, with minimal configuration, will learn how to operate the HVAC equipment more efficiently and how to collaborate with other actors of the grid. To this aim, we pursue to develop new Deep Learning and Reinforcement Learning methods, algorithms and tools to address three key issues: (1) simulation of buildings under different operations and contexts; (2) generation of optimal control instructions for HVAC to save energy while guaranteeing comfort; (3) coordination between components of the energy system to achieve an overall reduction of the contaminant emissions.

We are confident that our proposal could create in 5 years the technologies to reduce a 30% the energy consumed by buildings and increment a 30% the use of clean energies in grids, in line with current regulations. Not only do we expect a reduction on energy consumption and an increment on the use of renewable sources, but also a reduction on the cost of controlling energy in buildings, which is crucial to achieve a widespread adoption of climate mitigation technologies.


Interested? You can find more info here

Poster Paper

Research Team

Juan Gomez-Romero

Juan Gómez-Romero

Associate Professor
Universidad de Granada, Spain

Miguel Molina-Solana

Miguel Molina-Solana

Marie Curie Research Fellow
Universidad de Granada, Spain
Imperial College London, UK