This thesis shows the detailed design of the control and software of a DC microgrid of a tiny house community on the roof of a high-rise building in the city of Rotterdam in the Netherlands, consisting of twelve tiny houses powered by solar and wind energy. This thesis is part of a project with two other subgroups, focusing on the microgrid design and the powerline communication.

First, an introduction to the problem is given together with a description of the tiny house community. After that, the general program of requirements is presented, as well as the requirements of this subgroup. Next, an artificial neural network design is presented, which is used to forecast solar and wind generation and energy demand. The designed dense neural network resulted in predictions with mean errors of 10.11%, 12.56%, and 6.95% as a fraction of the maximum value for solar generation, wind generation, and energy demand, respectively. The predictions functioned as an input for the model predictive controller, which used them to place restrictions on appliances in the community when necessary, to reduce dependency on the main power grid of Rotterdam. Using a mathematical optimization algorithm, a simulation of one year showed that the controller could reduce the grid dependency up to 25%, compared to simulating without the controller. The conclusion summarises the achieved results, discusses whether the requirements are met, and considers possible future works.