Tackling Climate Change with next-generation Building Energy Simulation and Control

What is SAIL about?

SAIL aims at developing new Machine Learning technologies to progress towards climate-neutral buildings in their operation.

The overarching objective is to create novel ML technologies to materialize intelligent control in the future energy ecosystem. A smart energy grid controlled by autonomous self-adaptive agents that, with minimal and inexpensive configuration, will automatically learn how to operate the equipment more efficiently, and how to negotiate energy exchange with other actors.


Self-adaptive optimal and long-term control

Industrial control is usually carried out by a combination of reactive and rule-based controllers, which are difficult to configure, inefficient in systems with high inertia, and not adaptable to evolving environments.

Data-driven simulation models supporting automatic control.

Physical simulation models can be used to estimate system responses to alternative control sequences, and therefore to support optimization. However, developing such models is very time-consuming, and their execution time is very high, which makes them unsuitable for real-time control.

Game-theoretical approaches to confederated control.

Decentralization in confederated control opens a plethora of possibilities for agent interaction, ranging from collaboration to competition. Current control algorithms cannot deal with this problem without a significant amount of prior knowledge.

Transfer, multi-task, and meta-learning algorithms to extend simulation and control models to different scenarios.

AI-based control solutions cannot be reused from system to system, thus making it necessary to configure and retrain the models from scratch with a subsequent high effort.


Our vision is informed by the evidence and previous results we have obtained through different research projects.