Energy Optimal Control of a Multivalent Building Energy System using Machine Learning

In this contribution we develop and analyse intelligent control methods in order to optimise the energy efficiency of a modern residential building with multiple renewable energy sources. Because of alternative energy production options a non-convex mixed-integer optimisation problem arises. For the solution we first apply combined optimisation methods and integrate it into a model predictive controller. In comparison, a reinforcement learning based approach is developed and evaluated in detail. Both methods, in particular reinforcement learning approaches are able to decrease energy consumption and keep thermal comfort at the same time.