Minimizing the energy consumption of buildings while maintaining the comfort of the occupants is one key to decrease our global energy consumption. Recently, data-driven Machine Learning methods, typically based on neural networks, have emerged to take advantage of the growing amount of data available. In this talk, I will show issues of classical methods and a possible solution to still leverage the great expressiveness of neural networks for building control.

Download the slides for this talk.Download ( PDF, 3894.27 MB)