Islanded microgrids are small electric power networks that have no connection to a larger grid. They can be found, for example, in remote places, such as, islands or rural areas. Islanded microgrids typically comprise storage, renewable and conventional units as well as local loads. The central question of this work is: How to operate islanded microgrids with very high renewable share, i.e., how to control the energy of the storage units and how to maximize infeed from uncertain renewable sources without compromising a safe operation? To answer this question, different model predictive control (MPC) schemes for the operation of microgrids are deduced. These can be distinguished by the way they handle uncertain load and renewable infeed. Namely, they are: certainty equivalence MPC, minimax MPC, risk-neutral stochastic MPC, and risk-averse MPC. All schemes are posed in computationally tractable ways and compared in numerical case studies. These indicate that the way uncertain load and renewable infeed are modeled has a significant impact on safety and performance. Among the considered approaches, risk-averse MPC is most suitable as it provides robustness to misestimated forecasts and unlikely events which translates into low costs and a safe operation.weiterlesen